Information processing device

Abstract

The information processing apparatus of the present invention infers the road shape of the road on which its own vehicle is traveling based on map information obtained from a map information unit and sensor information obtained from a sensor unit. The information processing apparatus includes: a sensor resolution characteristic management unit that manages the sensor resolution characteristics of the sensor unit corresponding to the distance from the vehicle's position to the point where the intersection shape changes; an intersection appearance prediction unit that predicts the appearance of the intersection corresponding to the distance based on the sensor resolution characteristics; and an intersection model generation unit that generates an intersection model based on the intersection appearance predicted by the intersection appearance prediction unit.

Description

Technical Field

[0001] The present invention relates, for example, to an information processing device capable of accurately identifying roads surrounding a vehicle. Background Technology

[0002] The commercialization of preventative safety technologies in the automotive industry is entering a period of widespread adoption. While becoming increasingly multifunctional and high-performance, the scenarios they address are constantly expanding. These scenarios include, for example, complex environments such as ordinary roads and intersections where vehicles typically travel.

[0003] In recent years, the scenarios in which vehicles respond have expanded, and the ability of vehicle-mounted recognition devices to accurately identify intersections has become a significant challenge. However, existing technologies are primarily designed for vehicle control on a single road, so recognition devices rely solely on information from sensors such as cameras mounted on the vehicle to identify road shapes. However, conventional recognition devices struggle to properly identify complex road shapes such as intersections using only sensor data. Therefore, the technology disclosed in Patent Document 1 is considered to enable recognition devices to identify complex road shapes.

[0004] Patent document 1 describes "an identification device comprising: a road shape pattern selection unit that selects a road shape pattern based on the position of its own vehicle detected by its own vehicle position detection unit and road map information stored in a road map information storage unit; and a road shape recognition unit that recognizes the road shape in front of its own vehicle based on the detection information of the road shape information detection unit ahead and the road shape pattern selected by the road shape pattern selection unit".

[0005] Existing technical documents

[0006] Patent documents

[0007] Patent Document 1: Japanese Patent Application Publication No. 2000-30198 Summary of the Invention

[0008] The problem the invention aims to solve

[0009] In the identification device disclosed in Patent Document 1, the intersection ahead of the vehicle is modeled based on a map containing road structures, and the intersection model is fitted with sensor information to infer the intersection shape. However, this technology does not consider the following: due to the influence of sensor resolution, the appearance of the intersection will vary depending on the distance from the vehicle's position to the intersection. Especially for intersections located far from the vehicle, a mismatch will occur between the actual intersection shape and the intersection modeled by the identification device (hereinafter referred to as the "intersection model"). Therefore, the identification device's ability to infer the shape of the intersection is substantially limited to when the vehicle is approaching the intersection. Furthermore, the identification device cannot effectively utilize sensor information obtained from sensors observing intersections located far from the vehicle's position.

[0010] The present invention was made in view of the following situation, and its purpose is to correctly identify the shape of the road on which a vehicle travels.

[0011] Technical means to solve the problem

[0012] The information processing apparatus of the present invention infers the road shape of the road on which its own vehicle is traveling based on map information obtained from a map information unit and sensor information obtained from a sensor unit. The information processing apparatus includes: a sensor resolution characteristic management unit that manages the sensor resolution characteristics of the sensor unit corresponding to the distance from the vehicle's position to the point where the road shape changes; an appearance prediction unit that predicts the appearance of the road corresponding to the distance based on the sensor resolution characteristics obtained from the sensor resolution characteristic management unit; and a model generation unit that generates a road model based on the road appearance predicted by the appearance prediction unit.

[0013] The effects of the invention

[0014] According to the present invention, a road model is generated by predicting the appearance of the road corresponding to the distance from the vehicle based on the sensor resolution characteristics, so that the shape of the road on which the vehicle is traveling can be correctly identified.

[0015] Other issues, components, and effects not mentioned above will be clarified through the following description of the implementation method. Attached Figure Description

[0016] Figure 1 This is a block diagram illustrating an overall configuration example of an information processing apparatus according to an embodiment of the present invention.

[0017] Figure 2 This is a block diagram illustrating an example of the configuration of a sensor unit according to one embodiment of the present invention.

[0018] Figure 3 This is a block diagram illustrating an example of the configuration of a map information unit according to an embodiment of the present invention.

[0019] Figure 4 This is a block diagram illustrating an example of the configuration of an intersection information processing unit according to an embodiment of the present invention.

[0020] Figure 5 This is a block diagram illustrating an example of the configuration of a vehicle position determination unit according to an embodiment of the present invention.

[0021] Figure 6 This is a block diagram illustrating an example of the configuration of an intersection information parsing unit according to an embodiment of the present invention.

[0022] Figure 7 This is a block diagram illustrating an example of the configuration of a model supplementary information update unit according to an embodiment of the present invention.

[0023] Figure 8 This is a block diagram illustrating an example of the configuration of an intersection shape inference unit according to an embodiment of the present invention.

[0024] Figure 9 This is a block diagram illustrating an example of the configuration of an intersection consistency determination unit according to an embodiment of the present invention.

[0025] Figure 10 This figure illustrates an example of sensor information at a crossroads according to one embodiment of the present invention.

[0026] Figure 11 A diagram illustrating an example of the basic structure of an intersection according to one embodiment of the present invention.

[0027] Figure 12 This figure illustrates an example of a model selection unit switching between multiple models according to one embodiment of the present invention.

[0028] Figure 13 The figure illustrates an example of the basic structure of an intersection model according to one embodiment of the present invention and the detailed structure of the restored intersection model.

[0029] Figure 14A and Figure 14B A table showing examples of information regarding the road categories and road grades defined in the road structure order. Figure 14A A table listing the categories of roads, from type 1 to type 4, defined by each type of road and region. Figure 14B A table that records the planned traffic volume for each day, corresponding to the type of road and the terrain of the region.

[0030] Figure 15 This table represents examples of road category classifications and lane widths for ordinary roads as defined in the road structure order.

[0031] Figure 16A table showing examples of road category classifications and shoulder widths for ordinary roads as defined in the road structure order.

[0032] Figure 17A and Figure 17B A table showing examples of road information corresponding to road types and categories or road grades. Figure 17A This is a table that stores information on the number of lanes on one side, lane width, and roadside strips corresponding to the type and category of roads. Figure 17B This is a table that stores information on the number of lanes on one side, lane width, design speed, and roadside strips, corresponding to the road's grade.

[0033] Figure 18A and Figure 18B A table showing examples of road information corresponding to road types or speed limits. Figure 18A This is a table that stores information on the number of lanes on one side, lane width, design speed, and roadside strips corresponding to the type of road. Figure 18B This is a table that stores information on the number of lanes on one side, lane width, design speed, and roadside strips corresponding to the speed limit.

[0034] Figure 19 The diagram illustrates the detailed structure of an intersection according to one embodiment of the present invention, as well as an example of an intersection model.

[0035] Figure 20 The diagram illustrates examples of a first model and a second model generated based on the detailed structure of an intersection, according to one embodiment of the present invention.

[0036] Figure 21 The diagram illustrates examples of a first model and a second model generated for various types of intersections according to an embodiment of the present invention.

[0037] Figure 22 This figure illustrates an example of how the vehicle position determination unit of an embodiment of the present invention determines the correctness of its own vehicle position based on an intersection located near the front of the vehicle.

[0038] Figure 23 This diagram illustrates an example of the operation of a vehicle position detection unit according to an embodiment of the present invention.

[0039] Figure 24 This is a block diagram illustrating an example of the hardware configuration of a computer constituting an information processing apparatus according to an embodiment of the present invention. Detailed Implementation

[0040] [One Implementation Method]

[0041] Hereinafter, with reference to the accompanying drawings, an example of the configuration and operation of an information processing apparatus according to an embodiment of the present invention will be described. Figures 1-9 The accompanying drawings are for illustrating an example of the internal configuration of the information processing apparatus of this embodiment. Figure 10 The following are accompanying drawings illustrating the operation of the various functional units constituting the information processing device. Figures 14 to 18 show information such as road categories as stipulated in Japan's Road Structure Ordinance. In this specification and the accompanying drawings, constituent elements that substantially have the same function or structure are labeled with the same symbol, thereby omitting redundant explanations.

[0042] <Example of the overall structure of an information processing device>

[0043] First, refer to Figure 1 An example of the overall configuration of the information processing apparatus of this embodiment will be described.

[0044] Figure 1 This is a block diagram illustrating an example of the overall configuration of the information processing device 1. In this embodiment, the information processing device 1 is mounted in a vehicle (not shown). Hereinafter, the vehicle equipped with the information processing device 1 will be referred to as the "vehicle itself." The information processing device (information processing device 1) infers the road shape of the road on which its own vehicle is traveling based on map information obtained from the map information unit (map information unit 200) and sensor information obtained from the sensor unit (sensor unit 100). At this time, the information processing device 1 can effectively utilize sensor information obtained from onboard sensors (hereinafter referred to as "sensors") mounted in the vehicle and information from a map storing the road structure to identify roads and other elements surrounding the vehicle and determine its own vehicle's position. Furthermore, by effectively utilizing sensor information obtained from the sensors and a map storing the road structure, the information processing device 1 can also infer complex road shapes such as intersections. For example, in this embodiment, the road whose shape is inferred by the information processing device 1 is an intersection.

[0045] The information processing device 1 includes a sensor unit 100, a map information unit 200, an intersection information processing unit 300, an intersection shape inference unit 400, an intersection consistency determination unit 500, and a display-alarm-control unit 600.

[0046] The sensor unit 100 uses, for example, an onboard front sensor capable of observing objects or roads in front of the vehicle. In this embodiment, a stereo camera is referred to as an example of a front sensor, but the sensor could also be a monocular camera or a LiDAR (Laser Imaging Detection and Ranging) sensor. Furthermore, the sensor unit 100 outputs the information observed by the sensor in front of the vehicle as sensor information to the intersection information processing unit 300.

[0047] The map information unit 200 stores map information, which includes information about the roads the vehicle travels on (map, road type, road structure, etc.). The information stored by the map information unit 200 may be, for example, map information used by a car navigation system installed in the vehicle, or map information obtained from the Internet via wireless communication. This map information is stored, for example, on a high-capacity storage medium such as an HDD (Hard Disk Drive) or SSD (Solid State Drive) installed in the vehicle. Furthermore, the map information acquired by the map information unit 200 can represent simplified road shapes and also retains the function of storing and updating sensor information obtained by the sensor unit 100 within its observation range.

[0048] The intersection information processing unit 300 acquires map information for a specific area held by the map information unit 200. Next, the intersection information processing unit 300 analyzes the position and direction of its own vehicle within that area, and retrieves intersection information for the intersections ahead of the vehicle from the map information unit 200. Then, based on the acquired intersection information, the intersection information processing unit 300 uses default lane widths, shoulder positions, etc., to perform intersection information processing for generating an intersection model. An intersection model is a representation of the shape of an intersection using numbers or point groups.

[0049] If the map information includes information such as the number of lanes, the intersection information processing unit 300 uses this information to generate an intersection model. Alternatively, if the map information does not include information such as the number of lanes, the intersection information processing unit 300 determines the default number of lanes based on information such as the maximum vehicle speed in the map information, and then generates the intersection model. Furthermore, the intersection information processing unit 300 can also obtain intersection model information based on sensor information previously sensed by sensors.

[0050] The road shape inference unit (intersection shape inference unit 400) infers the road shape (intersection shape) based on the road model (intersection model) and sensor information, and outputs the inference result. For example, the intersection shape inference unit 400 uses sensor information obtained from the sensor unit 100 to perform inference processing for identifying the intersection shape on the intersection model generated by the intersection information processing unit 300. Therefore, the intersection shape inference unit 400 dynamically changes the road shape identification method and processing area, such as lanes and road ends, based on the intersection model predicted by the intersection information processing unit 300 and the error amount of the sensors used in the sensor unit 100, thereby identifying the intersection shape more stably and with higher accuracy.

[0051] The consistency determination unit (intersection consistency determination unit 500) determines the consistency between the sensor information obtained from the sensor unit 100 and the inference result obtained from the intersection shape inference unit 400, and outputs the determination result. Therefore, the intersection consistency determination unit 500 acquires a processing result including the intersection shape inferred by the intersection shape inference unit 400. Then, the intersection consistency determination unit 500 determines the consistency between the intersection model used by the intersection shape inference unit 400 in inferring the intersection shape and the sensor information obtained by the intersection shape inference unit 400 from the sensor unit 100, and outputs the determination result. At this time, the consistency determination unit (intersection consistency determination unit 500) uses the road model (intersection model) that infers the road shape (intersection shape) and accuracy information, and determines the consistency between the sensor information and the inference result by comparing the accuracy information with a predetermined threshold. Here, the accuracy information represents the degree of consistency between the sensor information and the road model (intersection model) when fitting the sensor information to the road model (intersection model), for example, expressed as confidence level. Therefore, if the accuracy information is higher than the specified threshold, the sensor information is sufficiently consistent with the road model (intersection model), and thus the determination result is obtained that the sensor information is consistent with the inference result obtained from the intersection shape inference unit 400.

[0052] The intersection information processing unit 300 generates an intersection model in front of its vehicle by referencing the map information unit 200 based on its own vehicle position. Therefore, if the vehicle's position is incorrect, the intersection information processing unit 300 may obtain intersection information from the map information unit 200 that differs from the intersection in front of its vehicle. For example, if the intersection information obtained from the map information unit 200 indicates a crossroads ahead of its vehicle, the intersection information processing unit 300 generates a crossroads-shaped intersection model. However, if the actual intersection in front of the vehicle is a T-shaped intersection, the intersection shape inference unit 400 may obtain sensor information from the sensor unit 100 that indicates a T-shaped intersection. Therefore, the intersection consistency determination unit 500 determines whether the generated intersection model contradicts the sensor information from the sensor unit 100. For example, if the intersection information obtained from the map information unit 200 is considered correct, the intersection consistency determination unit 500 may classify the sensor information from the sensor unit 100 as faulty. This determination process is called "consistency determination." Furthermore, there are instances where the map information managed by the Map Information Department 200 is outdated, while the sensor information actually observed by the Sensor Department 100 is newer, leading to inconsistencies between the map information and the sensor information. In such cases, the sensor information cannot be automatically deemed defective.

[0053] Here, a specific example of consistency determination is explained. The consistency determination unit (intersection consistency determination unit 500) compares the road model (intersection model) inferred by the road shape inference unit (intersection shape inference unit 400) with the road width and road structure connection relationships shown in the map information to determine the consistency between the sensor information and the inference result. For example, when the intersection model can be obtained from the intersection shape inference unit 400 as the inference result, the intersection consistency determination unit 500 performs consistency determination between the intersection model and the sensor information based on information such as the road width and road connection relationships of the obtained intersection model.

[0054] In addition, the intersection consistency determination unit 500 can also directly obtain the intersection model generated by the intersection information processing unit 300 and use the intersection model for consistency determination. Figure 1 (The connection pattern is shown by the dashed line). Furthermore, the intersection consistency determination unit 500 can also perform consistency determination by comparing the map information obtained from the map information unit 200 with the intersection model from which the intersection shape is inferred. Additionally, when a confidence level related to the inferred intersection shape can be obtained from the intersection shape inference unit 400, the intersection consistency determination unit 500 can also use the confidence level to perform consistency determination. Subsequently, the consistency determination unit (intersection consistency determination unit 500) stores the consistency determination results for intersections that its own vehicles have previously traversed.

[0055] The display-alarm-control unit 600 acquires the results from the intersection shape inference unit 400 and displays the recognition results. Furthermore, the display-alarm-control unit 600 displays driving support information to the driver of its own vehicle, issues warnings for safety assistance, or provides vehicle control and autonomous driving support. Additionally, the display-alarm-control unit 600 can also display and warn about the determination results from the intersection consistency determination unit 500. Therefore, the display-alarm-control unit 600 consists of a display device for displaying various information and a speaker for playing alarms.

[0056] <Example of the configuration of sensor unit 100>

[0057] Figure 2 This is a block diagram illustrating an example of the configuration of the sensor section 100.

[0058] The sensor unit 100 includes a left camera 101, a right camera 102, a matching unit 103, and a 3D point grouping unit 104.

[0059] As described above, the sensor unit 100 includes a vehicle-mounted front sensor. In this embodiment, a stereo camera consisting of a left camera 101 and a right camera 102 will be referred to as an example of a front sensor. Here, an embodiment of the stereo camera will be described. The sensing method for determining the distance to an object using a stereo camera is performed in the following order.

[0060] First, the left camera 101 and right camera 102, positioned at two different locations at the front of the vehicle, output left and right images captured within their respective fields of view. Next, if the same object is reflected in both the left and right images, the matching unit 103 determines the image position of the same object in both images through matching. After determining the image position of the same object, the matching unit 103 determines the difference in position reflected in the left and right images when both the left and right cameras 101 and 102 have captured the same object, thereby measuring the distance from the location of the stereo camera to the object.

[0061] The 3D point grouping unit 104 determines a triangle with the positions of the left camera 101 and the right camera 102 as the two apexes of the base of the triangle, and the position of the same object in the image as the vertex. It then reconstructs 3D points representing the three-dimensional position of the object using triangulation. Furthermore, the 3D point grouping unit 104 repeatedly performs 3D point reconstruction processing on the same object, thereby obtaining a large number of 3D point groups. Subsequently, the 3D point grouping unit 104 outputs the three-dimensional position of the object represented by the 3D point groups to the intersection shape inference unit 400.

[0062] <Example of the structure of Map Information Department 200>

[0063] Figure 3 A block diagram illustrating an example of the structure of the map information unit 200.

[0064] As described above, the map information unit 200 stores road-related map information and has the function of determining the location of its own vehicle on the map. The map information unit 200 of this embodiment includes a general map data unit 201, a GNSS (Global Navigation Satellite System) unit 202, a sensor detailed map storage unit 203, and a sensor detailed map update unit 204.

[0065] The General Map Data Unit 201 uses maps held by car navigation systems that represent road networks with nodes and connecting lines as general map data. However, the General Map Data Unit 201 is not limited to maps held by car navigation systems. For example, it can also be maps that are freely available on the Internet, or commercial maps that can be accessed via a server. As described below. Figure 11 As shown in the upper part, the general map data unit 201 consists of data composed of nodes representing locations on the map and connecting lines that link these nodes to form a road network. Therefore, the general map data unit 201 represents the road network based on this node and connecting line information.

[0066] The GNSS unit 202 uses GNSS information to determine the position of its own vehicle within the road network represented by node and link information in the general map data unit 201. Furthermore, this GNSS information can also be used to correct the vehicle's position using other sensors such as cameras, radar, gyroscopes, and its own vehicle behavior.

[0067] The sensor detail map storage unit 203 stores a map enhanced by information sensed by the sensor unit 100, such as lane width, road angle, number of lanes, distance from the outermost lane to the shoulder, and road shape—information not found in the general map—as sensor detail map data (referred to as "sensor detail map data"). Furthermore, the sensor detail map data updated in the sensor detail map storage unit 203 is added to the general map data managed by the general map data unit 201. The sensor unit 100 utilizes the general map data with the added sensor detail map data, thereby enabling it to perform sensing using the sensor detail map stored up to the previous trip when the vehicle passes the same road again. As a result, the sensor unit 100 can achieve more accurate and stable sensing.

[0068] The sensor detailed map update unit 204 has, for example, the following function: when the general map data is outdated or when roads are closed due to temporary construction, it updates the saved general map data using the sensor detailed map data. This update function is performed using sensor information output by the sensor unit 100. Through this update function, information such as new roads that are not present in the general map data can be added to the general map data. Furthermore, when a previously passable road becomes closed, the sensor detailed map data updated to indicate that the road is impassable is saved to the sensor detailed map storage unit 203, and then the general map data is updated.

[0069] However, if a vehicle only travels on the road once, the sensor detailed map update unit 204 will not decide to update the general map data, but will instead save it only as sensor detailed map data to the sensor detailed map storage unit 203. Therefore, after the sensor unit 100 senses that the road is impassable more than once, the sensor detailed map update unit 204 determines that the road is impassable and ultimately uses the sensor detailed map data to update the general map data. This judgment for update processing is the same for new roads that are not stored in the general map data. In other words, the sensor detailed map update unit 204 initially temporarily registers the information of new roads sensed by the sensor unit 100 in the sensor detailed map data, but does not register it as data in the final map. After confirming that the road is a reliable road by the vehicle passing through it multiple times, the sensor detailed map update unit 204 adds the sensor detailed map data containing the new road as data in the final map to the general map data.

[0070] <Example of the structure of an intersection information processing unit 300>

[0071] Figure 4 A block diagram illustrating an example of the configuration of the intersection information processing unit 300.

[0072] The intersection information processing unit 300 predicts the appearance of the intersection from the sensor viewpoint based on the intersection information existing in front of its vehicle obtained from the map information unit 200 and the sensor resolution characteristics corresponding to the sensor obtained from the sensor resolution characteristic storage unit 312, and generates an intersection model corresponding to the predicted appearance. This intersection information processing unit 300 includes a sensor resolution characteristic management unit 310, an intersection appearance prediction unit 320, and an intersection model generation unit 330.

[0073] (Example of the configuration of the sensor resolution characteristic management unit 310)

[0074] First, an example of the configuration of the sensor resolution characteristic management unit 310 will be explained.

[0075] The sensor resolution characteristic management unit (sensor resolution characteristic management unit 310) manages the sensor resolution characteristics of the sensor unit (sensor unit 100) corresponding to the distance from the vehicle's position to the point where the road shape changes. For example, the sensor resolution characteristic management unit 310 acquires the distance from the vehicle's position to the intersection and environmental information that affects the sensor's resolution characteristics, and outputs the sensor resolution characteristics corresponding to the acquired environmental information to the intersection appearance prediction unit 320. This sensor resolution characteristic management unit 310 includes an environmental information acquisition unit 311 and a sensor resolution characteristic storage unit 312.

[0076] The sensor resolution characteristic management unit (sensor resolution characteristic management unit 310) includes an environmental information acquisition unit (environmental information acquisition unit 311) that acquires environmental information affecting the sensor's observation limits. The environmental information acquisition unit 311 has the function of acquiring environmental information that affects the sensor's resolution characteristics and the sensor's observation limits. For example, in the case of sensors that observe brightness values, such as stereo cameras or monocular cameras, changes in ambient illumination will change the visibility of intersections. Thus, environmental conditions affect the ease or difficulty of observing objects, so the time related to changes in ambient illumination is acquired as environmental information.

[0077] Furthermore, weather conditions are also considered as a factor affecting changes in ambient illuminance, so the environmental information acquisition unit 311 can also acquire weather information as environmental information. Additionally, when the vehicle's headlights are on, the sensor's observation range may expand even at night or in low-light conditions such as inclement weather, so the environmental information acquisition unit 311 can also acquire the headlight's operating status as environmental information. Similarly, the illumination of streetlights around intersections also affects the sensor's observation range, so the environmental information acquisition unit 311 can also acquire the streetlight's operating status as environmental information. Furthermore, when there are obstacles such as other vehicles or pedestrians around the vehicle, the sensor cannot observe the intersection behind the obstacles, causing the intersection information processing unit 300 to be unable to acquire intersection information. Therefore, the environmental information acquisition unit 311 can also acquire information related to obstacles around the vehicle.

[0078] The sensor resolution characteristic management unit (sensor resolution characteristic management unit 310) includes a sensor resolution characteristic storage unit (sensor resolution characteristic storage unit 312) that stores the limit distance at which the sensor unit (sensor unit 100) can observe the unique shape of the road as a sensor resolution characteristic. For example, the sensor resolution characteristic storage unit 312 stores the distance at which the width of the road to the side of the intersection can be observed, the distance at which the curve of the lane in which the vehicle is traveling can be observed, etc., as sensor resolution characteristics. Furthermore, the sensor resolution characteristic management unit (sensor resolution characteristic management unit 310) reads the sensor resolution characteristics corresponding to the environmental information from the sensor resolution characteristic storage unit (sensor resolution characteristic storage unit 312) and outputs them to the appearance prediction unit (intersection appearance prediction unit 320). Here, see reference. Figure 10 An example of sensor information at a crossroads observed by sensor unit 100 will be explained.

[0079] <Example of sensor information at an intersection>

[0080] Figure 10This is a diagram illustrating an example of sensor information at an intersection. Marker 31, located near the center of each diagram, indicates the center position of the intersection (the point where two roads intersect) as sensed by the sensors.

[0081] (a) An intersection near one's own vehicle

[0082] When your vehicle is approaching the intersection, such Figure 10 As shown in the explanatory diagram (a), the sensor can observe the width of the road where the road branches at an intersection.

[0083] (b) An intersection located far from your vehicle

[0084] When the intersection is far away from your own vehicle, such as Figure 10 As shown in the explanatory diagram (b), the sensor cannot accurately observe the road width of the left and right forks. Therefore, the intersection information processing unit 300 cannot accurately determine the road width of the left and right forks based on the sensor information.

[0085] Therefore, the sensor resolution characteristic storage unit 312 stores the limit distance (hereinafter referred to as "first distance") of the road width on the side of the intersection where the sensor unit 100 can observe left and right forks as a sensor resolution characteristic. When the sensor resolution characteristic storage unit 312 can obtain road width information from the map information unit 200, the sensor resolution characteristic storage unit 312 can calculate the first distance by combining the sensor type of the sensor used to identify the shape of the intersection and road with the road width information obtained from the map information unit 200.

[0086] Furthermore, if the map information unit 200 contains information on the road category and grade as specified in the road structure order but not road width information, the first distance can be calculated by using the lane width and distance to the shoulder specified in the road structure order as road width information based on the category and grade information. Additionally, if road width information acquired during previous vehicle travel is stored, the sensor resolution characteristic storage unit 312 can use this road width information to calculate the first distance.

[0087] (c) Crossroads in low light

[0088] Furthermore, the sensor resolution characteristic storage unit 312 can also store the sensor's observation limit distance as a sensor resolution characteristic. In the case of sensors that observe brightness values, such as stereo cameras or monocular cameras, the sensor's observation range narrows in low-light environments. For example, the sensor's observation range narrows at night or in inclement weather, such as... Figure 10 As shown in the explanatory diagram (c), sensor information is obtained indicating that the sensor can only observe situations near its own vehicle. Therefore, it cannot observe the situation at intersections that are far away from its own vehicle.

[0089] Thus, the observation limit distance (hereinafter referred to as the "second distance") corresponding to changes in ambient illuminance has a significant impact on the appearance of the intersection, so the relationship between illuminance changes and the second distance can be stored in the sensor resolution characteristic storage unit 312. On the other hand, even in low-illuminance environments, the second distance will increase when the headlights are on or when streetlights near the intersection are lit. Therefore, the sensor resolution characteristic storage unit 312 can also store the second distance taking into account the operating status or lighting status of streetlights, and information indicating these conditions can also be stored together with the second distance.

[0090] (d) A crossroads partially blocked by an obstacle

[0091] Furthermore, the sensor resolution characteristic storage unit 312 can also store the position and size of obstacles acquired by the environmental information acquisition unit 311. For example, a car parked on the road near an intersection serves as an obstacle. In this case, such as Figure 10 As shown in the explanatory diagram (d), the sensor observes an intersection partially obscured by an obstacle. The obstruction range 32 of the intersection or road (the shaded area in the diagram) depends on the location and size of the obstacle. Therefore, the sensor resolution characteristic storage unit 312 can also store the obstruction range 32 or information indicating the location and size of the obstacle.

[0092] (Example of the configuration of the intersection appearance prediction unit 320)

[0093] Next, refer to Figure 4 The configuration example of the intersection appearance prediction unit 320 will be explained.

[0094] The appearance prediction unit (intersection appearance prediction unit 320) predicts the appearance of a road corresponding to a distance based on the sensor resolution characteristics obtained from the sensor resolution characteristic management unit (sensor resolution characteristic management unit 310). For example, the intersection appearance prediction unit 320 predicts the appearance of the intersection as observed by the sensor based on the sensor resolution characteristics input from the sensor resolution characteristic management unit 310. This intersection appearance prediction unit 320 includes its own vehicle position determination unit 321, intersection information parsing unit 322, and model selection unit 323.

[0095] (Example of the configuration of the vehicle position determination unit 321)

[0096] Next, refer to Figure 5 To illustrate the configuration of the vehicle position determination unit 321, we will explain its operation.

[0097] Figure 5 A block diagram illustrating an example of the configuration of the vehicle position determination unit 321.

[0098] The vehicle location determination unit (vehicle location determination unit 321) determines the vehicle's location on the map contained in the map information. The vehicle location determination unit 321 includes a GNSS error analysis unit 3211, a sensor error analysis unit 3212, and a time series comprehensive correction unit 3213.

[0099] The GNSS error analysis unit 3211 performs error analysis of the vehicle's position using GNSS information acquired through the vehicle's own behavior and a navigation system (not shown). Here, the GNSS error analysis unit 3211 analyzes the error (GNSS error) while comparing the vehicle's current location on the map with map information. The analysis result, i.e., error information, is used to correct the vehicle's position for higher accuracy. Therefore, the GNSS error analysis unit 3211 outputs the analyzed error information to the time series synthesis and correction unit 3213.

[0100] The sensor error analysis unit 3212 analyzes the longitudinal and lateral position errors (sensor errors) of its own vehicle position on the map using information obtained from sensors such as inertial sensors (gyroscopes), cameras, millimeter-wave sensors, and LiDAR. Then, the sensor error analysis unit 3212 outputs error information to the time series synthesis and correction unit 3213.

[0101] The time-series comprehensive correction unit 3213 uses error information obtained from the GNSS error analysis unit 3211 and the sensor error analysis unit 3212 to correct its own vehicle position on the map. However, if the time-series comprehensive correction unit 3213 corrects its own vehicle position based on GNSS errors or sensor errors according to instantaneous determination results, it is easy to update its own vehicle position with uncertain information or cause unstable position updates. Therefore, the time-series comprehensive correction unit 3213 analyzes errors in a time-series manner and integrates the analysis results to implement stable error correction.

[0102] (Example of the structure of the intersection information analysis unit 322)

[0103] Next, to explain the operation of the intersection information analysis unit 322, refer to... Figure 6 Let's illustrate the structure of the intersection information analysis unit 322.

[0104] Figure 6 A block diagram illustrating an example of the configuration of the intersection information parsing unit 322.

[0105] The road information parsing unit (intersection information parsing unit 322) acquires road shape information (intersection information) representing the shape of the road (intersection shape) existing ahead of the vehicle's direction of travel based on the determined position of the vehicle and map information, and parses the road shape information (intersection information). For example, the intersection information parsing unit 322 parses the intersection information acquired from the map information unit 200. This intersection information parsing unit 322 includes a road data reading unit 3221 and a basic shape parsing unit 3222. The intersection information parsing unit 322 first performs the reading of map information equivalent to the vehicle's road direction of travel in the road data reading unit 3221, and includes information described later. Figure 11 The data for reading the node and link information shown at the top is read in.

[0106] (Basic structure of an intersection)

[0107] Here, for reference Figure 11 The basic structure of the intersection is explained.

[0108] Figure 11 A diagram illustrating an example of the basic structure of an intersection. Figure 11 The upper part displays an example of a map and the vehicle's own location. Figure 11 The lower part shows examples of intersection shapes predicted based on the basic structures of Y-shaped, T-shaped, and crossroads.

[0109] Figure 11 The upper part of the map displays the approximate map area where the vehicle is located. This map area is, for example, divided into 5km squares. The travel road data reading unit 3221 then reads, for example, data from the map information unit 200 containing information about nodes and connecting lines within the 5km square area where the vehicle is located. The node and connecting line information data shown in the basic structure is used to identify roads (referred to as "travel roads") existing in the vehicle's direction of travel from the map.

[0110] Figure 6 The basic shape analysis unit 3222, as shown, obtains the basic structure of the intersections surrounding its own vehicle, composed of node and connecting line information, from the travel road data reading unit 3221. Furthermore, the basic shape analysis unit 3222 predicts the shape of the intersection existing in front of its own vehicle based on its own vehicle position and the analysis results related to its own travel road.

[0111] For example, in indicating that one's own vehicle is in Figure 11 When traveling upwards on the map shown at the top, the basic shape analysis unit 3222 predicts that the intersection in front of its own vehicle is a crossroads and the distance from its own vehicle position to the crossroads. Figure 11The lower part displays the basic shape analysis unit 3222, which selects the basic structure of the intersection in front of its vehicle as a cross from multiple intersections represented by basic structures. Here, to indicate that intersections not represented by a basic structure other than a cross are not selected, the basic structures of Y-shaped and T-shaped intersections other than crosses are grayed out.

[0112] (Example of the operation of model selection unit 323)

[0113] Next, return to Figure 4 The operation of the model selection unit 323 will be explained.

[0114] The model selection unit (model selection unit 323) selects the model that the model generation unit (intersection model generation unit 330) can generate based on the sensor resolution characteristics and sensor information obtained from the sensor resolution characteristic management unit (sensor resolution characteristic management unit 310) according to road shape information. For example, the model selection unit 323 selects whether to use a high-resolution model or a low-resolution model based on the intersection information of the intersection in front of its own vehicle obtained from the map information unit 200 and the sensor resolution characteristics obtained from the sensor resolution characteristic storage unit 312, and outputs the selection result to the intersection model generation unit 330. Here, the high-resolution model is the model used when the lateral road width can be detected by using the sensor information of the sensor unit 100, and is referred to as the "first model" in the following description. On the other hand, the low-resolution model is the model used when the lateral road width cannot be detected by using the sensor information of the sensor unit 100, and is referred to as the "second model" in the following description.

[0115] (Choose an example using one model)

[0116] The model selection unit 323 first obtains the intersection category ahead of its vehicle from the intersection information parsing unit 322. Next, the model selection unit 323 obtains the first distance managed in the sensor resolution characteristic management unit 310 based on the obtained intersection category. Then, the model selection unit 323 compares the distance from its own vehicle position to the intersection obtained from the intersection information parsing unit 322 with the first distance. Subsequently, if it determines that the distance from its own vehicle position to the intersection is shorter than the first distance and the sensor can observe the road width to the side of the intersection, the model selection unit 323 selects the first model. Conversely, if it determines that the distance from its own vehicle position to the intersection is larger than the first distance and the sensor cannot observe the road width to the side of the intersection, the model selection unit 323 selects the second model. Finally, the model selection unit 323 outputs the selection result to the intersection model generation unit 330.

[0117] (An example of switching between using multiple models)

[0118] Furthermore, the model selection unit 323 can switch between using more than three models by comparing the distance from its own vehicle position to the intersection with the first distance, instead of simply switching between the first model or the second model.

[0119] Figure 12 This diagram illustrates an example of the model selection unit 323 switching between using multiple models. Here, the variable model that can be treated as a single model is referred to as the "third model". When the model selection unit 323 selects the third model, the intersection model generation unit 330 generates multiple models that can be switched, as follows.

[0120] For example, if the intersection is far away relative to the first distance, the sensor cannot observe the road width on the side of the intersection. Therefore, the intersection model generation unit 330 uses... Figure 12 The model shown in the explanatory diagram (a) is generated in the form of a crossroads that is far away from the vehicle's own position relative to the first distance.

[0121] On the other hand, when the intersection is near the first distance, the width of the road to the side of the intersection observed by the sensor increases as the vehicle approaches the intersection. Furthermore, when the distance from the vehicle's position to the intersection is just below the first distance, the sensor can barely observe the width of the road to the side of the intersection. Therefore, the intersection model generation unit 330... Figure 12 The model shown in the explanatory diagram (b) generates intersections located at a distance less than the first distance from the vehicle's own position.

[0122] Furthermore, as a vehicle approaches an intersection, the road width on the side of the intersection widens. Figure 12 The left side of the explanatory diagram (c) shows the situation when the vehicle is approaching an intersection. The thick solid lines in the diagram represent the road boundaries forming the side lanes of the intersection. Using the vehicle's position in the diagram as the apex of a sector, the area within the sensor's field of view, represented by a single-dotted line, is shown as the area obstructed by obstacles and obstructed by dashed lines. Therefore, the intersection model generation unit 330 uses... Figure 12 The model shown on the right side of the explanatory diagram (c) generates an intersection near its own vehicle. This model indicates that the intersection of the roads that make up the intersection has become clearly defined.

[0123] The model selection unit 323 outputs the selection result of model 3 (selected as the third model) and the width of the road on the side of the intersection observed by the sensor to the intersection model generation unit 330. In the intersection model generation unit 330, the model is generated according to... Figure 12The diagrams (a), (b), and (c) illustrate the sequentially changing intersection models. Thus, when the third model is selected by the model selection unit 323, the intersection model generated by the intersection model generation unit 330 changes continuously as the vehicle approaches the intersection.

[0124] Furthermore, if the second distance can be obtained from the sensor resolution characteristic management unit 310, the model selection unit 323 can also output the obtained second distance along with the selection result of whether to use the first model or the second model to the intersection model generation unit 330. The intersection model generation unit 330 can generate the first model or the second model based on the second distance. Additionally, if obstacle information around the vehicle can be obtained from the sensor resolution characteristic management unit 310, the model selection unit 323 can also output this obstacle information to the intersection model generation unit 330. By using the obstacle information, the intersection model generation unit 330 understands that the sensor cannot accurately observe the area in front of the vehicle at the current time due to obstacles within its observation range.

[0125] (Example of the configuration of the intersection model generation unit 330)

[0126] Next, return to Figure 4 The composition and operation examples of the intersection model generation unit 330 are explained.

[0127] The model generation unit (intersection model generation unit 330) generates a road model (intersection model) by modeling the road based on the road appearance predicted by the appearance prediction unit (intersection appearance prediction unit 320). For example, the intersection model generation unit 330 acquires intersection information obtained from the intersection information parsing unit 322 and model information selected by the model selection unit 323, and generates an intersection model corresponding to the sensor resolution characteristics obtained from the sensor resolution characteristic management unit 310. This intersection model generation unit 330 includes a model supplementary information acquisition unit 331, a model supplementary information update unit 332, a high-resolution model generation unit 333, and a low-resolution model generation unit 334.

[0128] Here, for reference Figure 13 The intersection model generated by the intersection model generation unit 330 will be explained.

[0129] Figure 13 A diagram illustrating the basic structure of an intersection model and the detailed structure of the restored intersection model.

[0130] Figure 13 The upper side shows an example of a crossroads as a basic structure of an intersection. The basic structure of a crossroads is, for example, through... Figure 11 The basic shape analysis unit 3222 shown is predicted by the analysis process.

[0131] Figure 4 The intersection model generation unit 330 shown obtains information from the intersection information parsing unit 322. Figure 13 The basic structure of the intersection is shown on the upper side. The model additional information acquisition unit 331 acquires additional information (road width, angle, etc.) and attaches the additional information to the basic structure represented by node and link line information, thereby reconstructing the intersection. Figure 13 The detailed structure of the intersection is shown on the lower side.

[0132] Figure 13 The lower part shows the process of restoring the detailed structure of the intersection model by appending default values ​​or sensing values ​​acquired up to the last time to the basic structure of the intersection. Default values ​​or sensing values ​​acquired up to the last time include, for example, the number of lanes, lane width, shoulder position, and intersection angle. Thus, as... Figure 13 As shown on the lower side, for example, an intersection model is generated with a shoulder position of 1.6m, a lane width of 3.0m, 3 lanes on one side, and a shoulder intersection angle and a white line intersection angle of 90 degrees.

[0133] When the map information unit 200 stores additional information such as the number of lanes, lane width, and the angle between roads or white lines, the model additional information acquisition unit 331 acquires the additional information from the map information unit 200. However, depending on the map information used by the map information unit 200, sometimes detailed additional information such as the number of lanes, lane width, the angle between roads, and the distance from a lane to a shoulder cannot be acquired. When the model additional information acquisition unit 331 cannot acquire even some of this information, the intersection model generation unit 330 uses default information. This default information can be set to fixed information such as the number of lanes (2), lane width (3m), and shoulder width (1.5m), but it can also be set to default values ​​based on information such as the road category or grade in the map information unit 200 to generate model additional information.

[0134] (Example of default value)

[0135] Next, refer to Figure 14~ Figure 16 The default information, including default values, used by the intersection model generation unit 330 is explained.

[0136] As described above, when there is no additional model information in the map information unit 200, the intersection model generation unit 330 uses indirect information to set default values ​​corresponding to the road conditions. Compared to using fixed default values, by switching to default values ​​set according to the road conditions, the intersection model generation unit 330 can restore more detailed road shapes with higher accuracy for sensor observation.

[0137] Figure 14A and Figure 14BA table showing examples of information about the road categories and road grades defined in the road structure order.

[0138] Figure 14A This is a table that records the categories of roads, from type 1 to type 4, defined by road type and region.

[0139] Figure 14B A table that records the planned traffic volume (units / day) for each day, corresponding to the type of road and the terrain of the region.

[0140] Figure 14A , Figure 14B The information on the road category and grade is stored in the map information unit 200. In this case, the intersection model generation unit 330 can accurately reconstruct a detailed intersection model of the road shape by using the road category and grade information obtained from the map information unit 200 and the specified values ​​defined in the road structure order as default values.

[0141] Figure 15 This table represents examples of road category classifications and lane widths for ordinary roads as defined in the road structure order.

[0142] Road classification is based on road categories (Type 1 to Type 4), defining road levels (any one or more levels from 1 to 4). Furthermore, for each road category and level, a lane width is defined as a prescribed value for ordinary roads. Therefore, Figure 15 The road category classification and lane width information of ordinary roads are stored in the map information unit 200. In this case, the intersection model generation unit 330 can accurately reconstruct the intersection model of detailed road shapes by using the road classification and lane width of ordinary roads obtained from the map information unit 200 and applying the specified values ​​defined in the road structure order as default values.

[0143] Figure 16 A table showing examples of road category classifications and shoulder widths for ordinary roads as defined in the road structure order.

[0144] Road classification is based on road categories (Type 1 to Type 4), defining road levels (any one or more from Level 1 to Level 5). Furthermore, for each road category and level, the widths of the left and right shoulders of ordinary roads are defined as prescribed values. Therefore, Figure 16The road category classification and the width of the left and right shoulders of ordinary roads are stored in the map information unit 200. In this case, the intersection model generation unit 330 can accurately reconstruct an intersection model with detailed road shapes by using specified values ​​as default values ​​based on the road category classification and the width of the shoulders of ordinary roads obtained from the map information unit 200.

[0145] Additionally, Map Information Department 200 sometimes omits Figure 14. Figure 16 The information shown includes the road category and grade. Therefore, the intersection model generation unit 330 sometimes cannot reference the default values ​​defined by the map information unit 200 based on road category and grade information. Therefore, the intersection model generation unit 330 can also use other information held by the map information unit 200 to switch between default values ​​based on the road, thereby restoring an intersection model with a detailed road shape. This example will be explained with reference to Figures 17 and 18.

[0146] Figure 17A as well as Figure 17B A table showing examples of road information corresponding to road types and categories or road grades. Figure 17 and... Figure 17B The tables shown are examples of alternative solutions used when the intersection model generation unit 330 in the map information unit 200 does not have direct information such as lane width, number of lanes, and shoulder for generating intersection models, and therefore does not have any of the road category or grade information.

[0147] Figure 17A This is a table that stores information on the number of lanes on one side, lane width, and roadside strips corresponding to the type and category of roads. Figure 17A The table only contains road type and category information, not grade information. Relying solely on... Figure 17A The table cannot represent the full specifications of the detailed road structure using one-dimensional category information. However, Figure 17A The information stored in the table is useful for predicting the approximate width and number of lanes of a road. Therefore, compared to the case of no information at all, the intersection model generation unit 330 can accurately predict road information and reconstruct the intersection model.

[0148] Figure 17B This is a table that stores information on the number of lanes on one side, lane width, design speed, and roadside strips, corresponding to the road's grade. Figure 17B The table only contains road grade information, not category information. Relying solely on this... Figure 17B The table also cannot represent the full specifications of the detailed road structure using one-dimensional category information. However, Figure 17BThe information stored in the table is useful for predicting the approximate width and number of lanes of a road. Therefore, compared to the case of no information at all, the intersection model generation unit 330 can accurately predict road information and reconstruct the intersection model.

[0149] Figure 18A as well as Figure 18B A table showing examples of road information corresponding to road type or speed limit. Figure 18A as well as Figure 18B The tables shown are examples of alternative solutions used when the map information unit 200 does not have direct information such as lane width, number of lanes, and shoulder for generating intersection models using the intersection model generation unit 330, and therefore does not have information on road category and grade.

[0150] Figure 18A This is a table that stores information on the number of lanes on one side, lane width, design speed, and roadside strips corresponding to the type of road. Figure 18A The table does not contain information on road categories and grades, but it roughly stores the types of roads. In other words, Figure 18A The table contains default values ​​corresponding to the road types. Using this... Figure 18A The table uses default values ​​(fixed values) to appropriately set the information used by the intersection model generation unit 330 to generate the intersection model, based on the type of road the vehicle is traveling on. Furthermore, default values ​​for each detailed road shape are set in a table (not shown) for the intersection model generation unit 330 to reference. Moreover, even when information such as the number of lanes is partially stored in the general map, the intersection model generation unit 330 prioritizes using the information stored in the general map, only using the default values ​​corresponding to the situation when the information is missing, thereby striving for high accuracy in the intersection model.

[0151] Figure 18B This is a table that stores information on the number of lanes on one side, lane width, design speed, and roadside strips corresponding to the speed limit. Figure 18B The table does not contain information on road categories and grades, but it does have default values ​​corresponding to speed limits. (Using this...) Figure 18B The table uses default values ​​to appropriately set the information used by the intersection model generation unit 330 to generate the intersection model based on the speed limit of the road on which the vehicle is traveling. Furthermore, default values ​​for each detailed road shape are set in the table.

[0152] Furthermore, there are cases where the map information unit 200 does not store information related to road types but does store information related to speed limits, or where speed limit information is obtained through sensor information or communication. In these cases, a method based on... Figure 18BThe intersection model generation unit 330 can dynamically switch default values ​​based on the speed limit, as shown in the table. Alternatively, the intersection model generation unit 330 can indirectly and dynamically set the number of lanes, lane width, and roadside strip distance based on the road conditions on which its vehicles are traveling.

[0153] Furthermore, there are cases where the intersection model generation unit 330 obtains the road speed limit using information received by the map information unit 200 via communication, map information possessed by the map information unit 200, or sensor information obtained by the sensor unit 100, even if it cannot obtain the information based on the tables shown in Figures 14 to 18. In this case, the intersection model generation unit 330 can generate the intersection model using a table that references the default values ​​of the roads corresponding to the obtained speed limit. Further, if it cannot obtain any information from the information received via communication, map information, or sensor information, the intersection model generation unit 330 can generate the intersection model using completely fixed default values.

[0154] (Example of the configuration of the model additional information update unit 332)

[0155] Next, refer to Figure 7 ,right Figure 4 The configuration and operation examples of the model additional information update unit 332 shown will be explained.

[0156] Figure 7 A block diagram illustrating an example of the configuration of the model additional information update unit 332.

[0157] When a vehicle passes through the same location, the model-attached information update unit 332 determines whether the sensor information acquired up to the last time can be used based on the confidence level of the sensor information. In this determination, sensor information observed by the sensor unit 100, such as road ends and lanes, and road undulation information, is used. If the confidence level of the sensor information is below a certain threshold, the model-attached information update unit 332 uses this sensor information only as a provisional value for this time and does not send it to the map information unit 200 for data saving or updating. Conversely, if the sensor information has sufficient confidence above the threshold, the model-attached information update unit 332 sends the sensor information to the map information unit 200. The map information unit 200 then saves or updates the received sensor information.

[0158] However, if the map information unit 200 receives sensor information only once, it will not save or update the data using that sensor information. Therefore, when the map information unit 200 receives sensor information with high confidence multiple times, it will use the sensor information with high confidence to save or update the data. Alternatively, if the sensor information obtained by the sensor unit 100 in this observation has extremely high confidence and thus has high consistency with the saved data, the map information unit 200 can fuse the saved data with the current sensor information to update the map data.

[0159] The model supplementary information update unit 332 includes a road end update unit 3321, a lane update unit 3322, and an undulation update unit 3323. The road end, lane, and undulation are all used as useful model supplementary information for the intersection model generation unit 330 to generate a high-precision intersection model.

[0160] The road end updating unit 3321 updates information related to the positional relationship between the outermost lane and the road end. Furthermore, the road end updating unit 3321 also updates information such as changes in the shape of the lanes and road ends.

[0161] The lane update unit 3322 performs a judgment on the reuse of sensor information corresponding to the confidence level based on the number of lanes on the road, the complex changes in lane shape at intersections or merging points, lane width information, and the angles between roads, etc., which are not stored by the map information unit 200.

[0162] The undulation update unit 3323 processes information related to the unevenness and slope of the road. For example, for small or large convex bumps used for speed suppression on roads leading into residential areas, if there is a certain degree of confidence in the shape and position, the undulation update unit 3323 also determines whether to pass information related to the undulation of these roads to the map information unit 200.

[0163] Furthermore, when the sensor information from the sensor unit 100 exceeds a certain level of confidence, the intersection model generation unit 330 saves the sensor information to the map information unit 200. Moreover, the intersection model generation unit 330 reuses the sensor information saved in the map information unit 200, thus allowing the intersection model generation unit 330 to use prior knowledge, such as lane widths, corresponding to the road, before the sensor unit 100 performs a sensing action due to vehicle movement.

[0164] However, when a vehicle has only traveled a single route, it's difficult to ascertain whether the lane widths and other parameters of the road have been accurately obtained. For example, the lane width contained in the sensor information might be a provisional lane width under construction. Furthermore, it's possible that the sensor unit 100 might observe incorrect information, or that the white line might be obscured by a parked vehicle, resulting in a different lane width than initially perceived. In such a scenario, the intersection model generation unit 330, while confirming that the identified lane width during travel exceeds a certain level of confidence, repeatedly confirms that the same lane width was identified when the vehicle travels the same road, thereby continuously increasing the confidence level of the sensor information. This confidence level is, for example, added to the sensor information stored in the map information unit 200. Moreover, since sensor information is not stored in the map information unit 200 on roads that the vehicle has only traveled a single time, the intersection model generation unit 330 often identifies this sensor information as having low confidence.

[0165] After obtaining model supplementary information from the model supplementary information acquisition unit 331, the intersection model generation unit 330 generates an intersection model based on the selection result of the model selection unit 323. Here, the model generation unit (intersection model generation unit 330) has a high-resolution model generation unit (high-resolution model generation unit 333), which generates a first model corresponding to a specific road existing in front of its own vehicle position and observed by the sensor unit (sensor unit 100) when the distance is within a predetermined value. Furthermore, the model generation unit (intersection model generation unit 330) has a low-resolution model generation unit (low-resolution model generation unit 334), which generates a second model corresponding to a specific road that cannot be observed by the sensor unit (sensor unit 100) when the distance is greater than the predetermined value.

[0166] The intersection model is generated by either a high-resolution model generation unit 333 or a low-resolution model generation unit 334. When the sensor unit 100 observes the road width on the side of the intersection and the model selection unit 323 selects the first model, the high-resolution model generation unit 333 generates the intersection model. Conversely, when the sensor unit 100 cannot observe the road width on the side of the intersection and the model selection unit 323 selects the second model, the low-resolution model generation unit 334 generates the intersection model.

[0167] <Methods for representing intersection models>

[0168] Here, for reference Figure 19 The representation method of the intersection model generated by the intersection model generation unit 330 is explained.

[0169] Figure 19A diagram illustrating the detailed structure of an intersection and an example of an intersection model.

[0170] Figure 19 The top left shows an example of the detailed structure of the intersection. The detailed structure of this intersection is reconstructed by adding additional information obtained by the model additional information acquisition unit 331 to the basic structure represented by node and link information.

[0171] Figure 19 Model diagram (a) shows an example of a straight line model. The intersection model generation unit 330 uses the straight line model shown in model diagram (a) to represent the detailed structure of an intersection using straight lines. This straight line model is used for shape inference of intersections existing in residential areas and surrounded by walls or fences of houses.

[0172] Figure 19 Model diagram (b) shows an example of an arc model. When representing the detailed structure of an intersection using arcs, the intersection model generation unit 330 uses the arc model shown in model diagram (b). At intersections larger than those in residential areas, such as those on main roads, the corners of the intersection are designed as arcs. Therefore, the arc model is used for shape deduction of intersections where corners are designed as arcs.

[0173] Figure 19 Model diagram (c) shows an example of a polyline model. When using polylines to represent the detailed structure of an intersection, the polyline model shown in model diagram (c) is used. When the intersection model generation unit 330 wants to represent more detailed intersection shapes, such as curved mirrors or signs existing inside the intersection, or left and right turn lanes added near the intersection, a polyline model is used.

[0174] Figure 19 Model diagram (d) shows an example of a camera view model. The camera view model shown in model diagram (d) can be used when you want to represent an intersection as seen from a camera's perspective.

[0175] The intersection model generation unit 330 can obtain the intersection width from the model supplementary information acquisition unit 331. Figure 19 The intersection model generation unit selects one model from various models that represent the detailed structure of the intersection. For example, if the model additional information acquisition unit 331 infers that the road width is about one lane on each side, the intersection is likely to be located in a residential area and surrounded by walls or fences of houses, so the intersection model generation unit 330 selects a straight line model. Furthermore, if the model additional information acquisition unit 331 infers that the road width is two lanes or more on each side, the intersection is likely to be a major arterial road or an intersection larger than that in a residential area, so the intersection model generation unit 330 selects a circular arc model.

[0176] Furthermore, the intersection model generation unit 330 can select a model based on the application being used. For example, in applications using a locator function that misjudges its own vehicle position based on the consistency between map and sensor input, the intersection model generation unit 330 only needs to be able to infer the approximate shape of the intersection. Therefore, the intersection model generation unit 330 can select either a straight line model or a circular arc model to generate the intersection model. Additionally, in vehicle control applications where the vehicle can automatically turn left or right at intersections, the intersection model generation unit 330 can select a polyline model that can infer the detailed shape of the intersection to generate the intersection model.

[0177] <Methods for generating multiple intersection models>

[0178] Here, for reference Figure 20 The generation methods for the models generated by the high-resolution model generation unit 333 and the low-resolution model generation unit 334 are explained.

[0179] Figure 20 The diagram illustrates examples of the first and second models generated based on the detailed structure of the intersection. Figure 20 The top left corner shows the same as Figure 19 The top left shows an example of the detailed structure of the same intersection.

[0180] Figure 20 Model diagram (a) shows an example of the first model. The high-resolution model generation unit 333, based on... Figure 20 The detailed structure of the intersection shown in the upper left corner is used to generate the first model, which is used as an example of a high-resolution model. Figure 20 Model diagram (a) shows the high-resolution model generation unit 333. Figure 19 The model shown in Figure (a) represents the first model of the intersection using a straight-line model. However, the high-resolution model generation unit 333 can also use road width information or suitable applications. Figure 19 The model diagrams (b), (c), and (d) show different representation methods of the models.

[0181] Figure 20 Model diagram (b) shows an example of the second model. The low-resolution model generation unit 334, according to... Figure 20 The detailed structure of the intersection shown in the upper left corner is used to generate a second model, which serves as an example of a low-resolution model. Figure 20 Model diagram (b) shows the low-resolution model generation unit 334. Figure 19 The model shown in diagram (a) represents the second model of the intersection using a straight-line model. However, the low-resolution model generation unit 334 can also be used based on road width information or suitable applications. Figure 19 The model diagrams (b), (c), and (d) show different representation methods of the models.

[0182] <Examples of Model 1 and Model 2 corresponding to the intersection category>

[0183] Figure 20 The example shows the situation at a crossroads. For other intersection categories, the low-resolution model generation unit 334 can take into account the sensor resolution to generate the first and second models.

[0184] Figure 21 A diagram illustrating examples of the first and second models generated for each category of intersection.

[0185] Figure 21 Examples are shown of a first model generated by the high-resolution model generation unit 333 and a second model generated by the low-resolution model generation unit 334 for each type of intersection, such as a crossroads, a T-junction, and other intersection configurations. The first model is generated when vehicles are approaching the intersection, so the shape and detailed structure of the intersection are similar. On the other hand, the second model is generated when vehicles are far from the intersection, so the shape of the intersection is coarse. Therefore, by generating the second model when vehicles are far from the intersection and the first model when vehicles are approaching the intersection, an intersection model that closely resembles the detailed structure of the intersection is obtained.

[0186] <Intersection Shape Deduction Unit 400>

[0187] Next, refer to Figure 8 The configuration and operation of the intersection shape inference unit 400 will be explained.

[0188] Figure 8 A block diagram illustrating an example of the configuration of the intersection shape inference unit 400.

[0189] The road shape inference unit (intersection shape inference unit 400) outputs a road model (intersection model) inferred from the sensor information and the road model (intersection model) generated by the model generation unit (intersection model generation unit 330), along with accuracy information related to the inference result. Therefore, the intersection shape inference unit 400 fits the intersection model generated by the intersection model generation unit 330 to the features of the intersection shown by the sensor information obtained from the sensor unit 100. At this time, the intersection shape inference unit 400 infers the parameters constituting the intersection model (e.g., the performance characteristics described later). Figure 19 (The two-dimensional coordinates of the nodes in the model shown). The intersection shape inference unit 400 includes a road end recognition unit 410, a lane recognition unit 420, and a parameter inference unit 430.

[0190] (Example of the operation of the roadside identification unit 410)

[0191] Here, an example of the operation of the roadside identification unit 410 will be explained.

[0192] The intersection shape inference unit 400 uses the road end recognition unit 410 to obtain road end features for model fitting of the intersection. Road end features are information obtained by the road end recognition unit 410 from sensor information, extracting features of steps that are higher or lower in three dimensions relative to the road. Therefore, in cases where there is a road shoulder relative to the road, and further, a wall to the side of the shoulder, the road end recognition unit 410 extracts features of both the shoulder and the wall from the sensor information. Furthermore, the road end recognition unit 410 also extracts various three-dimensional objects such as walls, buildings, trees, utility poles, and ditches from the sensor information as features.

[0193] For a vehicle traveling on the road, the important step is the one that it will first encounter from the vehicle's perspective. Therefore, it is important for the vehicle to travel in a way that avoids contact with the step. Thus, the roadside recognition unit 410 emphasizes steps that are likely to be encountered first from the perspective of the vehicle, and performs noise reduction processing such as rejection for three-dimensional objects that exist at a position farther away from the vehicle.

[0194] After performing these processes, the road end recognition unit 410 uses the steps that its own vehicle is likely to first encounter as the road end point group, centered on the vehicle itself. Furthermore, the road end recognition unit 410 uses this road end point group for its own vehicle behavior, or uses corresponding points from the image to construct the road end point group in a time-series manner. Then, the parameter inference unit 430 fits this time-series road end point group with the intersection model generated by the intersection model generation unit 330 to infer the parameters of the intersection model.

[0195] (Example of the operation of lane recognition unit 420)

[0196] Next, an example of the operation of the lane recognition unit 420 will be explained.

[0197] The intersection shape inference unit 400 can also use the lane features identified by the lane recognition unit 420 to fit the intersection model, similar to the process described above which uses the recognition results of the road end recognition unit 410 to infer the intersection shape. The lane recognition unit 420 detects a group of points that are characteristic of the boundary lines between lanes or the boundary lines between lanes and shoulders along its own vehicle's direction of travel. Furthermore, the lane recognition unit 420 also uses its own vehicle's behavior to generate lane features that continuously connect the detected point groups in a time series manner, and identifies lanes based on these lane features. The parameter inference unit 430 fits the generated lane features to the intersection model, thereby inferring the parameters of the intersection model.

[0198] (Example of the operation of parameter inference unit 430)

[0199] Next, an example of the operation of the parameter inference unit 430 will be explained.

[0200] The parameter inference unit 430 acquires roadside features from the roadside recognition unit 410 and lane features from the lane recognition unit 420. Then, the parameter inference unit 430 calculates the parameters of the intersection model that best fits these acquired features. Figure 19 (The two-dimensional coordinates of the node shown). Subsequently, the intersection shape inference unit 400 images... Figure 1 As shown, the parameters of the inferred intersection model and the model selection information (which model to choose, i.e., the first, second, or third model) are output to the intersection consistency determination unit 500. Furthermore, if the degree of fit between the intersection model and the sensor information can be calculated, the intersection shape inference unit 400 also outputs the degree of fit to the intersection consistency determination unit 500.

[0201] Furthermore, the processing performed in the intersection shape inference unit 400 sometimes includes inferring the intersection shape without using information from the map information unit 200. In this case, the parameter inference unit 430 can skip parameter inference, and the intersection model generation unit 330 can directly output the intersection model it generates along with the intersection shape inferred by the sensor unit 100 to the intersection consistency determination unit 500.

[0202] <Example of the structure and operation of the intersection consistency determination unit 500>

[0203] Next, refer to Figure 9 The composition and operation examples of the intersection consistency determination unit 500 are explained.

[0204] Figure 9 A block diagram illustrating an example of the configuration of the intersection consistency determination unit 500.

[0205] The consistency determination unit (intersection consistency determination unit 500) determines the consistency of multiple roads (intersections) existing ahead of its own vehicle's direction of travel. The intersection consistency determination unit 500 includes a confidence determination unit 510, a vehicle position determination unit 520, a vehicle position exploration unit 530, and a vehicle position storage unit 540.

[0206] (Example of the operation of the confidence determination unit 510)

[0207] The confidence determination unit 510 determines the consistency between the intersection model used in the intersection shape inference and the sensor information obtained from the sensor unit 100 based on the inference result input from the intersection shape inference unit 400.

[0208] When a first model of the intersection shape inferred by the intersection shape inference unit 400 is available, the confidence determination unit 510 performs a consistency determination based on the road width of the obtained first model. For example, the case where the intersection shape inferred by the intersection shape inference unit 400 is a crossroads where two (two lanes) roads intersect will be explained.

[0209] For example, regarding the width of the four roads present at an intersection (two roads in the direction of the vehicle's travel and two roads in the opposite direction, and two roads to the right and left that intersect with these two roads at the intersection), the confidence determination unit 510 assumes that there may be roads narrower than the width of its own vehicle. In this case, it is assumed that the model used by the intersection shape inference unit 400 in inferring the shape of the intersection is incorrect or that the sensor information obtained from the sensor unit 100 contains a large amount of noise, so the model fits the features of non-intersections. Therefore, the confidence determination unit 510 determines that the consistency between the intersection model and the sensor information has not been obtained. Furthermore, regarding the four roads present at the intersection, when the confidence determination unit 510 compares the width of the two roads that the vehicle passes through while traveling straight through the intersection, there may be cases where the width obtained from the intersection model is significantly different from the actual width inferred from the sensor information. In this case, the confidence determination unit 510 may determine that the consistency between the intersection model and the sensor information has not been obtained.

[0210] Furthermore, when a second model is obtained after the intersection shape is inferred by the intersection shape inference unit 400, the confidence determination unit 510 performs a consistency determination based on the road width of the obtained second model, similar to the determination process performed using the first model. However, like the reference... Figure 20 As previously explained, the second model represents the intersection using two straight lines. Therefore, the confidence determination unit 510 determines consistency by comparing the distance between the two straight lines with the width of its own vehicle.

[0211] Furthermore, when a third model is obtained after the intersection shape inferred by the intersection shape inference unit 400, the confidence determination unit 510 performs a consistency determination of the first model if the distance from its own vehicle position to the intersection is smaller than the first distance. And, if the distance from its own vehicle position to the intersection is larger than the first distance, the confidence determination unit 510 performs a consistency determination similar to that of the second model.

[0212] Furthermore, the confidence determination unit 510 can also determine the consistency between the map information obtained from the map information unit 200 and the intersection model after the intersection shape inference unit 400 infers the intersection shape. When the map information unit 200 indicates that there is an intersection in front of its vehicle, the connection angles of the four roads at the intersection are right angles. Therefore, if the connection angles of the intersection model inferred by the intersection shape inference unit 400 deviate significantly from right angles, the confidence determination unit 510 can determine that consistency has not been achieved.

[0213] Furthermore, the map information unit 200 sometimes contains information that can infer road width (road width information, road category and grade in the road structure order, road width inference results from the last trip, etc.). Therefore, the confidence determination unit 510 can also determine the consistency between the road width information and the intersection model by comparing the road width information obtained from the road width inference information read from the map information unit 200 with the intersection model inferred by the intersection shape inference unit 400.

[0214] Furthermore, if the confidence level related to the inference result can be obtained from the intersection shape inference unit 400, the confidence level determination unit 510 can set a threshold for the confidence level to determine consistency. Additionally, if the intersection model generated by the intersection model generation unit 330 and the intersection shape inferred by the sensor unit 100 can be obtained from the intersection shape inference unit 400, the confidence level determination unit 510 can determine the consistency between the intersection model and the intersection shape by comparing the two.

[0215] (Example of the operation of the vehicle position determination unit 520)

[0216] Next, an example of the operation of the vehicle position determination unit 520 will be explained.

[0217] In this embodiment, the intersection model generation unit 330 generates an intersection model in front of its vehicle based on the vehicle's position obtained by the intersection information processing unit 300 and with reference to the map information unit 200. Then, the intersection shape inference unit 400 infers the intersection shape. Therefore, if the vehicle's position on the map is incorrect, the intersection model generation unit 330 may obtain different intersection information from the map information unit 200, resulting in an incorrect intersection model. Therefore, the confidence determination unit 510 determines the consistency between the intersection model used in the intersection shape inference and the sensor information obtained from the sensor unit 100, thereby determining the accuracy of the vehicle's position.

[0218] Furthermore, the amount of information required for consistency determination by the intersection consistency determination unit 500 differs depending on whether the intersection shape inference in the intersection shape inference unit 400 uses the first model or the second model. For example, in the case of the first model, the vehicle is closer to the intersection than the first distance, so the intersection shape inference unit 400 can perform shape inference that includes the road width on the side of the intersection. Therefore, if the confidence determination unit 510 determines, for example, that the intersection model and the sensor information are consistent, the intersection consistency determination unit 500 can determine that there is an intersection in front of the vehicle.

[0219] On the other hand, in the case of the second model, it is impossible to infer the shape by including the road width on the side of the intersection. Therefore, as Figure 21 As shown, the exact same intersection model is used for both crossroads and T-junctions. If the exact same intersection model is used for both crossroads and T-junctions, even if the confidence determination unit 510 determines that the crossroads model and sensor information are consistent, it may still be unable to distinguish between a crossroads and a T-junction. Therefore, the confidence determination unit 510 cannot determine that the intersection in front of its vehicle is a crossroads.

[0220] Furthermore, the same second model is applied to the three categories of road shapes: T-junctions, corners, and end-of-the-road intersections. Therefore, even if the intersection is T-shaped, the same second model as for corners and end-of-the-road intersections is selected, so the confidence determination unit 510 cannot determine that the intersection in front of its vehicle is a T-junction. This inability of the confidence determination unit 510 to determine the intersection in front of its vehicle also applies to the relationship between crossroads and T-junctions.

[0221] However, if the sensor information obtained from the sensor unit 100, which is output by the sensor observing the area in front of the vehicle at the current time, is used, the confidence determination unit 510 can correctly determine that the intersection far from the vehicle is a crossroads. Conversely, even when a second model of a crossroads is obtained, the shape of the crossroads in the second model is different from that of a T-junction, a corner, or a road end. Therefore, the shape of the second model generated when the confidence determination unit 510 determines that the intersection far from the vehicle is a crossroads does not match the shape of the second model generated when the road shape is a T-junction, a corner, or a road end. Therefore, the confidence determination unit 510 can limit the shape candidates of the distant intersection.

[0222] Therefore, the consistency determination unit (intersection consistency determination unit 500) includes a vehicle position determination unit (vehicle position determination unit 520) that determines the correctness of its own vehicle position based on the consistency determination result of the road and outputs the determination result of its own vehicle position. For example, the vehicle position determination unit 520 obtains the type information of the selected model (one of the first model, the second model, or the third model) from the confidence determination unit 510 and the consistency determination result given by the intersection consistency determination unit 500, determines the correctness of its own vehicle position, and outputs the determination result. A detailed example of the processing performed in the vehicle position determination unit 520 will be explained below.

[0223] First, if the confidence determination unit 510 obtains a result indicating that the first model was used for shape inference and that the first model and sensor information are consistent, the vehicle position determination unit 520 outputs a determination that the vehicle's position is correct. Conversely, if the confidence determination unit 510 obtains a result indicating that the first model was used for shape inference and that the first model and sensor information are not consistent, the vehicle position determination unit 520 outputs a determination that the vehicle's position is incorrect.

[0224] Furthermore, if the confidence determination unit 510 obtains the result that the second model was used for shape inference and that the second model is consistent with the sensor information, the vehicle position determination unit 520 considers the vehicle's position to be correct at the current time. Therefore, the vehicle position determination unit 520 outputs a determination result stating that the vehicle's position is correct. However, there is a possibility that the vehicle position determination unit 520 may not be able to determine if the vehicle's position is correct at the current time. In this case, the vehicle position determination unit 520 may output a determination result stating that the vehicle's position is indeterminate.

[0225] Furthermore, consider a scenario where the vehicle position determination unit 520 obtains information from the confidence determination unit 510 that the third model is used for shape inference. In this case, the vehicle position determination unit 520 determines the vehicle position of the first model if the distance from the vehicle position to the intersection is smaller than the first distance, and if the distance is larger than the first distance, it makes a consistent determination similar to the vehicle position determination of the second model.

[0226] Here, for reference Figure 22 An example of the vehicle position determination unit 520 performing the correctness determination of its own vehicle position will be explained.

[0227] Figure 22This diagram illustrates an example of how the vehicle position determination unit 520 processes the determination of the correctness of its own vehicle position based on an intersection located in front of or near the vehicle. The diagram shows the actual position of the vehicle (hereinafter referred to as "first vehicle position") and the position of the vehicle incorrectly inferred by GNSS (hereinafter referred to as "second vehicle position").

[0228] Figure 22 The explanatory diagram (a) shown on the left illustrates an example of the processing when the vehicle position determination unit 520 does not use the information of distant intersections in the locator. Not using the information of distant intersections in the locator means that the vehicle position determination unit 520 does not use the intersection information of intersections located far from the vehicle to determine the vehicle's position on the map.

[0229] In the diagram, the first vehicle position is indicated by vehicle icon 51, and the second vehicle position is indicated by vehicle icon 52. An intersection 61 is located near the first vehicle position, and an intersection 63 is located further ahead. Furthermore, an intersection 62 is shown near the second vehicle position, and a corner 64 is shown further ahead. Further, vehicle icon 51 shows the range 56 that the information processing device 1 can infer the intersection model. Range 56 is approximately equal to the range that the sensor unit 100 can observe. Moreover, the range 56, including the intersection 61 located in front of the vehicle shown at the first vehicle position, is observed by the sensor unit 100.

[0230] When there are multiple intersections of the same type (such as crossroads) around the vehicle shown on the map, the vehicle location determination unit 520 may sometimes misjudge the vehicle's location. Figure 22 In the example shown, there are intersections 61 and 62 on the map in front of vehicle icons 51 and 52. Here, it is assumed that the intersection model generation unit 330 generates an intersection model based on the intersection 62 in front of the second vehicle position (vehicle icon 52). In this case, since there is also an intersection 61 in front of the first vehicle position (vehicle icon 51), the confidence determination unit 510 determines that the generated intersection model is consistent with the sensor information.

[0231] As a result, although the vehicle's actual position is the first position, the vehicle position determination unit 520 determines that the second position is correct. Thus, in Figure (a), information about distant intersections 63 and 64 was not used by the locator, so the vehicle only realized its positional error when it approached intersection 63 after passing the first intersection 61 and proceeding in a straight line. Therefore, without using information about distant intersections in the locator and when similar intersections (such as crossroads) exist around the vehicle, it is difficult to determine the vehicle's position.

[0232] Next, refer to Figure 22 The explanatory diagram (b) shown on the right explains the countermeasures for the vehicle position determination unit 520 misjudging the vehicle's position. Here, the vehicle position determination unit 520 uses intersection information from a distant intersection as a locator and determines the correctness of its own vehicle position based on the consistency of multiple intersections existing in front of the vehicle. "Using the locator" refers to the vehicle position determination unit 520 using intersection information to determine the correctness of its own vehicle position on the map.

[0233] Explanatory Figure (b) shows a situation where there is an intersection less than a first distance from the vehicle in front of the first vehicle position (vehicle icon 51), and also an intersection at a distance greater than a first distance from the vehicle. Furthermore, Explanatory Figure (b) shows a situation where there is an intersection 62 less than a first distance from the vehicle in front of the second vehicle position (vehicle icon 52), and a corner 64 at a distance greater than a first distance from the vehicle. The model inference range 57, which the information processing device 1 can infer from the intersection model, is shown for vehicle icon 51. Moreover, the model inference range 57, including intersections 61 and 63 in front of the vehicle, is observed by the sensor unit 100.

[0234] Here, we will discuss the case where the intersection model generation unit 330 generates an intersection model at the second position of its own vehicle as shown by vehicle icon 52. In this case, the intersection model generation unit 330 generates an intersection model 63a as the first model based on the intersection 62 that exists in front from its own vehicle's view, and generates a corner model 64a as the second model based on the corner 64 that is far away from its own vehicle's view.

[0235] On the other hand, since there is an intersection 61 near the first vehicle position (vehicle icon 51) where the vehicle actually exists, the confidence determination unit 510 determines that the intersection of the first model is consistent. However, there is an intersection 63 far from the first vehicle position.

[0236] If the vehicle position determination unit 520 determines the vehicle's position on the map as the second vehicle position (vehicle icon 52), the shape of the corner model 64a generated by the intersection model generation unit 330 based on the corner 64 is different from the shape of the intersection model 63a generated by the intersection 63. Therefore, the confidence determination unit 510 determines the consistency between the corner model 64a (which did not obtain the second model) and the shape of the intersection 63 observed by the sensor unit 100 within the model inference range 57. As a result, the vehicle position determination unit 520 can determine that the vehicle's position is incorrect. Thus, the vehicle position determination unit 520 can determine the correctness of the vehicle's position by confirming the consistency between the intersection information of the consecutive intersections in front of the vehicle and the map and sensor information.

[0237] Figure 22 The example shown illustrates how the vehicle position determination unit 520 determines its own vehicle position using two intersections: an intersection 61 that is nearby relative to the first distance and an intersection 63 that is far relative to the first distance. However, the intersections used by the vehicle position determination unit 520 for determining its own vehicle position can be two or more intersections that are nearby relative to the first distance, and two or more intersections that are far relative to the first distance. Furthermore, two or more intersections can all be nearby relative to the first distance, and two or more intersections can all be far relative to the first distance. Additionally, the vehicle position determination unit 520 can also retain the results from the confidence determination unit 510, which determines the position based on multiple intersections the vehicle has previously passed, and combine this past time-series information to determine the accuracy of the vehicle position determination.

[0238] (Example of the operation of the vehicle's own position exploration unit 530)

[0239] Here, if the vehicle's own position determination unit 520 determines that the vehicle's position on the map is incorrect, Figure 9 The location shown is the most likely current location of the vehicle on the exploration map of the exploration department 530 (hereinafter referred to as "the third vehicle location").

[0240] Here, for reference Figure 23 The actions of the vehicle's position exploration unit 530 will be explained.

[0241] Figure 23 This diagram illustrates the actions of the vehicle's position exploration unit 530.

[0242] like Figure 23As shown in the explanatory diagram (a), a T-junction 72 exists in front of the second vehicle position (vehicle icon 52). Therefore, a first model or a second model of the T-junction 72 is generated based on the distance from the second vehicle position to the intersection. At this time, when sensor information indicating that the sensor unit 100 has observed an intersection is input, the vehicle position determination unit 520 determines that the shape of the intersection determined using the sensor information 82 is different from the shape of the intersection model 81. Therefore, the vehicle position determination unit 520 outputs the information that the second vehicle position is incorrect to the vehicle position exploration unit 530.

[0243] The consistency determination unit (intersection consistency determination unit 500) includes a vehicle position exploration unit (vehicle position exploration unit 530). When the vehicle position determination unit 520 determines the vehicle position determined on the map shown in the map information to be incorrect, the vehicle position exploration unit 530 explores the correct vehicle position on the map based on the road consistency determination result. Therefore, when the vehicle position exploration unit 530 receives information from the vehicle position determination unit 520 that the vehicle position is incorrect, it explores intersections adjacent to intersection 72 located ahead of the second vehicle position on the map, thereby determining a candidate vehicle position, i.e., a third vehicle position. Here, the vehicle position exploration unit 530 identifies intersections within a distance that the vehicle can traverse in one step (e.g., one zone on the map) as intersections adjacent to intersection 72 in a graphical structure displaying road relationships represented by nodes and edges.

[0244] like Figure 23 As shown in the explanatory diagram (a), among the intersections adjacent to the intersection 72 located in front of the second vehicle's position, there is one T-junction 73 and two crossroads 71 ​​and 74. Therefore, the intersection model generation unit 330 generates a first model or a second model based on the distance from the second vehicle's position to each intersection. Subsequently, the intersection consistency determination unit 500 determines the consistency between the generated first model or second model and the sensor information.

[0245] Here, the intersection models generated by the intersection model generation unit 330 are consistent with the sensor information at two intersections 71 and 74. Therefore, the vehicle position exploration unit 530 determines the positions near the intersections 71 and 74, which are consistent with the sensor information, as the third vehicle positions 53a and 53b. Subsequently, the vehicle position exploration unit 530 outputs the two third vehicle positions 53a and 53b shown in Figure (a) to the display-alarm-control unit 600.

[0246] Furthermore, the vehicle location exploration unit 530 can output the third vehicle location only if the third vehicle location is determined to be only one, and output the information that the vehicle location is undetermined if there are multiple third vehicle locations. In addition, the vehicle location exploration unit 530 can also save the information output to the display-alarm-control unit 600 to the vehicle location storage unit 540.

[0247] Figure 23 Explanatory diagram (b) shows the situation where the vehicle position exploration unit 530 uses the second model to explore the position of the vehicle itself. When the second model is selected by the model selection unit 323, the intersection model generation unit 330, for example, like... Figure 23 As shown, an intersection model 83 is generated that treats both the T-junction 75 and the corner 76 as having the same shape. Even if the intersections in front of the vehicle on the map are of different types (T-junction 75 and corner 76), as long as the shape of the intersection shown in the sensor information 84 is the same as the intersection model 83, the vehicle position exploration unit 530 will output two third vehicle positions 53c and 53d as candidates for the vehicle position.

[0248] Here, although not shown in the diagram, when the distance from the actual vehicle position to the intersection is less than the first distance, either the output result of the third vehicle position 53c or 53d can no longer represent the correct third vehicle position. Furthermore, when the intersection ahead of the first vehicle position is the end of a road, and the distance from the vehicle position to the intersection is less than the first distance, these output results are no longer the third vehicle position. Therefore, when the second model is selected, the vehicle position exploration unit 530 can output the information of the third vehicle position inferred using the second model, instead of only outputting the third vehicle position.

[0249] Furthermore, the intersections explored by the vehicle's own location exploration unit 530 from around its own location are not limited to those adjacent to the intersections located in front of the second vehicle location. For example, intersections within a distance that the vehicle can traverse in two or more steps can also be explored on a graphical structure displaying road relationships represented by nodes and edges. The number of steps used in this exploration can be determined by a default value or dynamically based on the GNSS position inference accuracy.

[0250] Furthermore, the vehicle location exploration unit 530 can also select intersections within a circle centered on the second vehicle location, instead of exploring intersections by the number of steps in the graphical structure. The radius of the circle can be determined by a default value or dynamically based on the GNSS position inference accuracy.

[0251] Furthermore, the vehicle position detection unit 530 can also detect the position of the third vehicle from multiple intersections located in front of the second vehicle position. Similar to the vehicle position determination unit 520, which determines the correctness of the vehicle position based on the consistency of multiple intersections, the vehicle position detection unit 530 also detects the third vehicle position based on information from multiple intersections. Through this operation, even when there are many intersections of the same type around the vehicle, the vehicle position detection unit 530 can more accurately infer the position of the third vehicle.

[0252] Furthermore, not limited to intersections in front of its own vehicle, the vehicle position exploration unit 530 can also retain the results of the intersection shape inference unit 400 and the confidence determination unit 510 from multiple intersections it has passed in the past, thereby combining past time series information to explore the third vehicle position.

[0253] (Example of the operation of the vehicle location storage unit 540)

[0254] Figure 9 The vehicle location storage unit 540 shown can store the first vehicle location, the second vehicle location determined by the vehicle location determination unit 520, the third vehicle location explored by the vehicle location exploration unit 530, and the past time sequence information of the vehicle's travel history. Therefore, as long as it is known from information such as GNSS that the road the vehicle is currently traveling on is near the road the vehicle has traveled on in the past, the intersection consistency determination unit 500 can use the vehicle location information read from the vehicle location storage unit 540 to determine the consistency of the intersection.

[0255] <Hardware Composition of Information Processing Devices>

[0256] Next, the hardware configuration of the computer 90 constituting the information processing device 1 will be described.

[0257] Figure 24 This is a block diagram illustrating an example of the hardware configuration of a computer 90. The computer 90 is an example of hardware used as a computer, which can operate as an information processing device 1.

[0258] The computer 90 includes a CPU (Central Processing Unit) 91, a ROM (Read-Only Memory) 92, and a RAM (Random Access Memory) 93, all connected to a bus 94. Furthermore, the computer 90 includes a display device 95, an input device 96, a non-volatile storage unit 97, and a network interface 98.

[0259] The CPU 91 reads the program code of the software implementing the functions of this embodiment from the ROM 92 and loads it into the RAM 93 for execution. Variables and parameters generated during the CPU 91's operation are temporarily written into the RAM 93, and the CPU 91 reads these variables and parameters as needed. However, an MPU (Micro Processing Unit) can also be used instead of the CPU 91. The processing of each functional unit in the information processing device 1 is performed by the CPU 91.

[0260] The display device 95, for example, is a liquid crystal display monitor, which displays the results of processing performed in the computer 90 to the driver. The input device 96, for example, uses a keyboard or mouse, allowing the driver to perform prescribed inputs and instructions. Alternatively, the display device 95 and the input device 96 may be integrated into a touch panel display. The display device 95 and the input device 96 correspond to... Figure 1 The display-alarm-control unit 600 is shown. Additionally... Figure 1 The information processing device 1 may be configured without the display-alarm-control unit 600, but it may also be configured to have the display-alarm-control unit 600.

[0261] The non-volatile storage unit 97 can be, for example, an HDD, SSD, floppy disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape, or other non-volatile memory. In addition to the operating system (OS) and various parameters, the non-volatile storage unit 97 also records programs used to enable the computer 90 to function. The ROM 92 and the non-volatile storage unit 97 record programs and data required for the CPU 91 to operate, and are used as an example of a computer-readable non-temporary storage medium storing programs executed by the computer 90. Various information and data generated within the information processing device 1 are saved to the non-volatile storage unit 97. Furthermore, Figure 1 The information processing device 1 may be configured without a map information unit 200, but it may also be configured to include a map information unit 200.

[0262] The network interface 98 may be a NIC (Network Interface Card), for example. The information processing device 1 communicates wirelessly via the network interface 98, and can upload data to or download data from an external server via the Internet. Furthermore, the information processing device 1 can access an in-vehicle network (e.g., CAN (Controller Area Network)) via the network interface 98 to obtain sensor information from the sensor unit 100 or output its own vehicle position information to a navigation system or autonomous driving system (not shown).

[0263] In the information processing apparatus 1 of the above-described embodiment, sensor information from intersections near the vehicle can be used not only, but also sensor information from intersections far from the vehicle to eliminate mismatches with the intersection model. In this case, the information processing apparatus 1 generates an intersection model by predicting the appearance of the road corresponding to the distance based on the sensor resolution characteristics, thus correctly identifying the shape of the intersection where the vehicle is traveling.

[0264] Furthermore, the information processing device 1 can accurately predict the appearance of the intersection generated from sensor information based on the sensor resolution characteristics that change according to the environment around the intersection and the environment around its own vehicle. In this way, the information processing device 1 can correctly identify the shape of the road on which its vehicle is traveling, so even in ordinary roads or complex scenes around intersections, it can output the correct road shape. Therefore, the driver or autonomous driving device can use the recognition results to ensure the safe driving of its vehicle.

[0265] Furthermore, the sensor resolution of the sensor unit 100 varies depending on the current environment of the vehicle. Therefore, the environmental information acquisition unit 311 acquires environmental information, and the sensor resolution characteristic storage unit 312 stores the sensor resolution characteristics that match the environmental information. Thus, for example, in situations where the location is the same but the time is different, the intersection information processing unit 300 can generate different intersection models based on the sensor resolution characteristics.

[0266] Furthermore, the intersection appearance prediction unit 320 predicts the appearance of the intersection based on its own vehicle position determined on the map by its own vehicle position determination unit 321 and the intersection information parsed by the intersection information parsing unit 322. At this time, the model selection unit 323 selects the intersection model generated by the intersection model generation unit 330. As the intersection model, it selects a first model as a high-resolution model, a second model as a low-resolution model, or switches between multiple models according to a time series. Subsequently, the intersection model generation unit 330 generates an intersection model based on the model selected by the model selection unit 323. Therefore, an appropriate model is selected based on the distance from the vehicle's own position to the intersection, which is then effectively used in the processing of the intersection model generation unit 330.

[0267] Furthermore, the intersection model generation unit 330 can accurately reconstruct an intersection model with detailed road shapes based on map information obtained from the map information unit 200, by using predefined values ​​defined in the road structure order as default values. Therefore, the information processing device 1 can correctly identify which section of the road its vehicle is located in, enabling, for example, autonomous driving or navigation systems to guide its vehicle's position with high precision.

[0268] Furthermore, the intersection shape inference unit 400 fits the intersection model generated by the intersection model generation unit 330 with the features of the intersection shown by the sensor information. Thus, the actual features of the intersection are appended to the intersection model, and the intersection shape inference unit 400 can infer the parameters of the intersection model in detail.

[0269] Furthermore, the intersection consistency determination unit 500 determines the consistency between the intersection model used in inferring the intersection shape and the sensor information, thereby correctly determining the shape of the intersection existing in front of its own vehicle.

[0270] Furthermore, the intersection consistency determination unit 500 uses information about intersections located far from its own vehicle in the locator, thereby enabling the correct determination of the vehicle's position when multiple intersections of the same shape exist around the vehicle, which were incorrectly determined.

[0271] [Variation Example]

[0272] Furthermore, if the sensor information of the sensor unit 100 contains information that can reproduce the shape of the intersection, then the sensor information of the sensor unit 100 can be input to the intersection consistency determination unit 500 without going through the intersection shape inference unit 400. Figure 1 (The connection pattern is shown by the dashed line). In this case, in the information processing device 1, the intersection model is input from the intersection information processing unit 300 to the intersection consistency determination unit 500. Therefore, the intersection consistency determination unit 500 can also determine the consistency of the intersection shown in the map information based on the intersection model and the sensor information input from the sensor unit 100. However, the information processing device 1 may also choose to perform the consistency determination of the intersection model and sensor information, and the consistency determination of the intersection model whose shape is inferred by the intersection shape inference unit 400 with the intersection shown in the map information, without going through the intersection shape inference unit 400.

[0273] Furthermore, in the above embodiment, the information processing device 1 performs appearance prediction, intersection model generation, and intersection shape inference for intersections in the direction of its own vehicle's travel. However, it is not limited to intersections. For example, in cases where the road is curved without branching or where there is a track crossing the road, the information processing device 1 can also perform appearance prediction, road model generation, and road shape inference for roads. In this case, tracks crossing the road and roads whose angle changes when viewed from the direction of travel of its own vehicle are identified as specific roads. Thus, curved roads and roads crossed by tracks also become objects of road shape inference in this embodiment.

[0274] Furthermore, the information processing device 1 described in the above embodiments can be a vehicle-mounted device installed in a vehicle, or a terminal device that can be removed from the vehicle and taken away.

[0275] Furthermore, the present invention is not limited to the above-described embodiments. As long as it does not depart from the spirit of the invention as described in the claims, various other applications and modifications can be adopted.

[0276] For example, the above embodiments are detailed and specific descriptions of the system configuration provided to illustrate the present invention in an easily understandable manner, and are not necessarily limited to having all the configurations described. Furthermore, other configurations may be added, deleted, or replaced in part of the configuration of this embodiment.

[0277] Furthermore, the control lines and information lines shown are only those deemed necessary for the instructions; not all control lines and information lines may be displayed on the actual product. In reality, almost all components can be considered interconnected.

[0278] Symbol Explanation

[0279] 1…Information processing unit, 100…Sensor unit, 200…Map information unit, 300…Intersection information processing unit, 310…Sensor resolution characteristic management unit, 320…Intersection appearance prediction unit, 330…Intersection model generation unit, 400…Intersection shape inference unit, 410…Road end recognition unit, 420…Lane recognition unit, 430…Parameter inference unit, 500…Intersection consistency determination unit, 510…Confidence determination unit, 520…Vehicle position determination unit, 530…Vehicle position exploration unit, 540…Vehicle position storage unit, 600…Display-alarm-control unit.

Claims

1. An information processing apparatus that infers the shape of an intersection on a road in which its own vehicle is traveling based on map information obtained from a map information unit and sensor information obtained from a sensor unit, characterized in that it comprises: The sensor resolution characteristic management unit manages the sensor resolution characteristics of the sensor unit corresponding to the distance from the vehicle's own position to the intersection. An appearance prediction unit predicts the appearance of the intersection corresponding to the distance based on the sensor resolution characteristics obtained from the sensor resolution characteristic management unit; and The model generation unit generates an intersection model based on the appearance of the intersection predicted by the appearance prediction unit, thus modeling the intersection. The sensor resolution characteristic management unit has a sensor resolution characteristic storage unit, which stores the limit distance at which the sensor unit can observe the unique shape of the intersection as the sensor resolution characteristic. The sensor resolution characteristic management unit includes an environmental information acquisition unit that acquires environmental information that affects the limiting distance. The sensor resolution characteristic management unit reads the sensor resolution characteristic corresponding to the environmental information from the sensor resolution characteristic storage unit and outputs it to the appearance prediction unit. The model generation unit has a high-resolution model generation unit and a low-resolution model generation unit. When the distance is within a specified value, the high-resolution model generation unit generates a first model corresponding to a specific road existing in front of the vehicle's position and observed by the sensor unit. When the distance is greater than the specified value, the low-resolution model generation unit generates a second model corresponding to the specific road that the sensor unit cannot observe.

2. The information processing device according to claim 1, characterized in that, The appearance prediction unit has: The vehicle location determination unit determines the vehicle's location on the map contained in the map information; The intersection information analysis unit obtains intersection information representing the shape of the intersection existing in front of the vehicle's direction of travel based on the determined position of the vehicle and the map information, and analyzes the intersection information. as well as The model selection unit selects a model that the model generation unit can generate based on the sensor resolution characteristics obtained from the sensor resolution characteristic management unit according to the intersection information and the sensor information.

3. The information processing device according to claim 1, characterized in that, The system includes an intersection shape inference unit, which infers the shape of the intersection based on the intersection model and the sensor information, and outputs the inference result.

4. The information processing apparatus according to claim 3, characterized in that, The intersection shape inference unit outputs the intersection model that infers the intersection shape and the accuracy information related to the inference result, based on the sensor information and the intersection model generated by the model generation unit.

5. The information processing apparatus according to claim 4, characterized in that, The system includes a consistency determination unit, which determines the consistency between the sensor information obtained from the sensor unit and the inference result obtained from the intersection shape inference unit, and outputs the determination result.

6. The information processing apparatus according to claim 5, characterized in that, The consistency determination unit uses the intersection model that infers the intersection shape and the accuracy information to determine the consistency between the sensor information and the inference result by comparing the accuracy information with a predetermined threshold.

7. The information processing apparatus according to claim 5, characterized in that, The consistency determination unit compares the intersection model inferred by the intersection shape inference unit with the road width and road structure connection relationship shown in the map information, thereby determining the consistency between the sensor information and the inference result.

8. The information processing apparatus according to claim 5, characterized in that, The consistency determination unit determines the consistency of the multiple intersections existing in the direction of travel of the vehicle itself.

9. The information processing apparatus according to claim 5, characterized in that, The consistency determination unit stores the consistency determination results of the intersections that its own vehicle has previously passed.

10. The information processing apparatus according to claim 5, characterized in that, The consistency determination unit has a vehicle position determination unit. The vehicle position determination unit determines the correctness of its own vehicle position based on the consistency determination result of the intersection, and outputs the determination result of its own vehicle position.

11. The information processing apparatus according to claim 10, characterized in that, The consistency determination unit has a vehicle location exploration unit. If the vehicle location determination unit determines that the vehicle location determined on the map shown in the map information is incorrect, the vehicle location exploration unit explores the correct vehicle location on the map based on the consistency determination result of the intersection.

Images