Blind angle area estimation device, travel environment generation device, vehicle control device, and vehicle

Abstract

The invention relates to a blind angle area estimation device, a driving environment generation device, a vehicle control device and a vehicle. A blind angle area of an observation area of the vehicle is estimated. The present invention is provided with: a travel route generation unit for calculating a travel route of a vehicle; an observation region switching position calculation unit that calculates an observation region switching position at which the observation region is switched from the first observation region to the second observation region when the travel path intersects the movement path of the other moving body; an observation region calculation unit that calculates a second observation region; a script generation unit that generates a plurality of evaluation scripts for different environmental conditions for the travel route; a recognition determination unit that simulates the sensing of the second observation area by the in-vehicle sensor in accordance with each evaluation script, and determines whether or not the recognition of another moving body in the second observation area is successful; and a risk map generation unit that generates a risk map in which a blind spot region in the second observation region and the second observation region are superimposed by integrating recognition success / failure results for the plurality of evaluation scripts.

Description

Technical Field

[0001] The present invention relates to a blind spot area estimation device, a driving environment generation device, a vehicle control device, and a vehicle. Background Art

[0002] In autonomous driving, sensors mounted on a vehicle sense an observation area that needs to be observed from the vehicle to confirm the presence of other vehicles or the like, and control the vehicle based on data acquired by the sensors. Therefore, in order for the sensors mounted on the vehicle to properly sense the observation area of the vehicle, it is necessary to appropriately set the area detected by the sensors.

[0003] As a technique for setting the detection area of sensors mounted on a vehicle, Patent Document 1 is known. In Patent Document 1, it is described in paragraph 0031 that: "The radar device 11 operates after switching the operation mode to either a field-of-view angle priority mode or a distance priority mode according to an instruction from the ECU 13. The field-of-view angle priority mode is a mode in which an object is detected in a detection area SA1 where the maximum detection distance is relatively short and the horizontal field-of-view angle is relatively wide, for example, as Figure 2 described. The distance priority mode is a mode in which an object is detected in a detection area SA2 where the maximum detection distance is relatively long and the horizontal field-of-view angle is relatively narrow, for example, as Figure 3 described."

[0004] Furthermore, regarding the processing performed by the ECU 13 that controls the radar device 11, it is described in paragraph 0041 that: "When the ECU 13 determines that there is an intersection or a turn ahead of the own vehicle 100, the process proceeds to step S3. On the other hand, when the ECU 13 determines that there is no intersection or turn ahead of the own vehicle 100, the process proceeds to step S5." It is described in paragraph 0042 that: "In step S3, the ECU 13 determines whether the position of another vehicle is acquired.... When the ECU 13 determines that the position of another vehicle is acquired, the process proceeds to step S4. On the other hand, when the ECU 13 determines that the position of another vehicle is not acquired, the process proceeds to step S5." In addition, it is described in paragraph 0043 that: "In step S4, the ECU 13 sets the radar device 11 to the distance priority mode." It is described in paragraph 0044 that: "In step S5, the ECU 13 sets the radar device 11 to the field-of-view angle priority mode."

[0005] Prior Art Documents

[0006] Patent Documents

[0007] Patent Document 1: Japanese Unexamined Patent Application Publication No. 2011 - 253241 Summary of the Invention

[0008] Even when the detection area of a sensor is set as in Patent Document 1, dead zones may sometimes be formed within the detection area of the sensor due to weather or time of day. Moreover, when the dead zone in the detection area of the sensor overlaps with the observation area of the vehicle, a dead zone is also formed in the observation area of the vehicle. If such dead zones can be estimated, by installing infrastructure sensors in the environment to sense dead zones and the like, the dead zones can be complemented, which helps improve the safety of autonomous driving.

[0009] Therefore, in the present invention, an object is to provide a dead zone estimation device, a driving environment generation device, a vehicle control device, and a vehicle that estimate dead zones in the observation area of a vehicle.

[0010] The dead zone estimation device of the present invention is, for example, a dead zone estimation device that estimates dead zones in the observation area of a vehicle that sets an observation area and performs autonomous driving based on data obtained by sensing the observation area with a sensor, and includes: a driving path generation unit that calculates a driving path to the destination of the vehicle; an observation area switching position calculation unit that, when the driving path intersects with the moving path of another moving body, calculates an observation area switching position for switching the observation area from a first observation area to a second observation area; an observation area calculation unit that calculates the second observation area corresponding to the observation area switching position; a scenario generation unit that generates a plurality of evaluation scenarios with different specified environmental conditions for the driving path; an identification determination unit that, for each of the evaluation scenarios, simulates sensing of the second observation area by a sensor mounted on the vehicle under the environmental conditions and determines whether or not the other moving body located within the second observation area can be successfully identified; and a risk map generation unit that integrates the results of whether or not the other moving body in the second observation area can be successfully identified for the plurality of evaluation scenarios and generates a risk map that overlays and shows the dead zones in the second observation area.

[0011] In addition, the driving environment generation device of the present invention is, for example, a driving environment generation device that estimates a dead angle area in the above-described observation area of a vehicle that sets an observation area and performs autonomous driving based on data obtained by sensing the above-described observation area, and has: a driving path generation unit that calculates a driving path to the destination of the vehicle; an observation area switching position calculation unit that calculates an observation area switching position for switching the above-described observation area from a first observation area to a second observation area when the above-described driving path intersects with the moving path of another moving body; an observation area calculation unit that calculates the above-described second observation area corresponding to the above-described observation area switching position; a risk map generation unit that generates a risk map showing the observation state of the surrounding environment based on the vehicle based on sensing data obtained from sensors mounted on the vehicle; and a dead angle area drawing unit that overlays and displays the above-described risk map and the above-described second observation area.

[0012] In addition, the vehicle control device of the present invention controls the vehicle in accordance with, for example, the above-described driving path, the above-described observation area switching position, and the above-described observation area calculated by the above-described dead angle area estimation device.

[0013] In addition, the vehicle of the present invention has, for example, the above-described vehicle control device.

[0014] Advantages of the Invention

[0015] According to the present invention, it is possible to provide a dead angle area estimation device, a driving environment generation device, a vehicle control device, and a vehicle that estimate a dead angle area in the observation area of a vehicle. Other problems and new features will become clear from the description of this specification and the drawings. Description of the Drawings

[0016] Figure 1 It is a diagram showing the structure of the dead angle area estimation device of Embodiment 1.

[0017] Figure 2 It is a diagram explaining road information.

[0018] Figure 3 It is a diagram explaining a three-dimensional map.

[0019] Figure 4 It is a diagram showing a flowchart of the processing performed by the dead angle area estimation device.

[0020] Figure 5 It is a diagram showing the flowchart in step S104.

[0021] Figure 6 It is a diagram showing an example of road information and a driving path.

[0022] Figure 7It is a diagram showing the second observation area when turning right at an intersection.

[0023] Figure 8 It is a diagram showing the second observation area when turning left at an intersection.

[0024] Figure 9 It is a diagram showing the second observation area when turning right at a T-junction without traffic lights.

[0025] Figure 10 It is a diagram showing the second observation area when turning left at a T-junction without traffic lights.

[0026] Figure 11 It is a diagram showing the flowchart in step S105.

[0027] Figure 12 It is a diagram showing the flowchart in step S106.

[0028] Figure 13A It is a diagram showing the first observation area switching position and the second observation area when turning right at an intersection.

[0029] Figure 13B It is a diagram showing Figure 13A An example of a risk map corresponding to the second observation area shown.

[0030] Figure 14A It is a diagram showing the second observation area switching position and the second observation area when turning right at an intersection for the second time.

[0031] Figure 14B It is a diagram showing Figure 14A An example of a risk map corresponding to the second observation area shown.

[0032] Figure 14C It is a diagram showing an example of a risk map updated based on an evaluation script with the same driving path but different weather conditions.

[0033] Figure 15 It is a diagram showing the structure of the blind spot area estimation device according to the second embodiment.

[0034] Figure 16 It is a diagram showing an example of the sensor information of the infrastructure sensor.

[0035] Figure 17 It is a diagram showing the flowchart of the processing performed by the infrastructure sensor configuration calculation unit.

[0036] Figure 18A It is a diagram showing the driving path of the host vehicle when turning right at an intersection.

[0037] Figure 18BThis is a diagram showing an example of a risk map generated for the Figure 18A driving route shown in

[0038] Figure 19 This is a diagram showing the structure of the driving environment generation device according to Embodiment 3.

[0039] Figure 20 This is a diagram showing an example of the drawing by the dead angle area drawing unit.

[0040] Explanation of Reference Numerals

[0041] 1 Dead angle area estimation device

[0042] 2 Driving environment generation device

[0043] 20 Destination acquisition unit

[0044] 21 Road information acquisition unit

[0045] 22 Three-dimensional map acquisition unit

[0046] 23 ODD condition acquisition unit

[0047] 30 Driving route generation unit

[0048] 40 Observation area determination unit

[0049] 41 Observation area switching position calculation unit

[0050] 42 Observation area calculation unit

[0051] 50 Script generation unit

[0052] 51 Path deviation amount estimation unit

[0053] 52 Evaluation script generation unit

[0054] 61 Identification judgment unit

[0055] 62 Risk map generation unit

[0056] 70 External function Detailed Implementation Modes

[0057] Hereinafter, embodiments of the present invention will be described with reference to the Figure One accompanying drawings.

[0058] Embodiment 1

[0059] Figure 1 This is a diagram showing the structure of the dead angle area estimation device 1 according to Embodiment 1. Figure 1The dead angle area estimation device 1 shown estimates the dead angle area in the observation area of a vehicle that sets an observation area and performs autonomous driving based on data obtained by sensing the observation area with a sensor. For example, the dead angle area estimation device 1 is mounted on a computer, reads in the ODD conditions including destination information, road information, 3D map, and vehicle information of the autonomous driving vehicle recorded in a storage device, and estimates the dead angle area. Further, a risk map representing the dead angle area is generated, and the generated risk map is output to an external function 70. Hereinafter, the autonomous driving vehicle that is the object of dead angle area estimation is referred to as the own vehicle.

[0060] The dead angle area estimation device 1 includes: a destination acquisition unit 20, a road information acquisition unit 21, a 3D map acquisition unit 22, an ODD condition acquisition unit 23, a driving path generation unit 30, an observation area determination unit 40, a script generation unit 50, an identification and determination unit 61, and a risk map generation unit 62.

[0061] The dead angle area estimation device 1 has, for example, a CPU, a GPU, a RAM, a ROM, etc., and can implement these functional units by expanding a prescribed program stored in the ROM into the RAM and executing it by the CPU. In addition, a part or all of the functions of the dead angle area estimation device 1 can also be implemented using hardware such as an FPGA or an ASIC.

[0062] The destination acquisition unit 20 acquires destination information and outputs it to the driving path generation unit 30. The destination information is information indicating the position to which the autonomous driving vehicle travels, and is expressed by, for example, latitude, longitude, and altitude. In the present embodiment, it is assumed that the destination acquisition unit 20 acquires the destination information by accepting the input of the destination information, but the destination information may also be stored in a storage medium or the like in advance and read in. In addition, the destination acquisition unit 20, for example, manages by associating the information of place names or buildings with latitude, longitude, and altitude in advance, and can acquire latitude, longitude, and altitude from the input information even when the information of place names or buildings is input.

[0063] The road information acquisition unit 21 acquires road information and outputs it to the driving path generation unit 30, the observation area determination unit 40, and the script generation unit 50. It is assumed that the road information acquisition unit 21 acquires the road information by reading in the road information stored in a storage medium or the like in advance, but it is not limited thereto. Figure 2 This is a diagram for explaining road information. As Figure 2 shown, the road information includes road network information, etc., and this road network information includes an edge 102 showing the moving path (driving path) of the vehicle and nodes 101 connecting the edges 102 to each other. In addition, the road information also includes information on lane areas, sidewalk areas, information on road markings, information on ground objects such as road signs 109 and traffic lights 110. Further,Figure 2 This is an example of road information, which is the road network information of a T-junction where a one-lane road on one side merges with a two-lane road on one side. The position information of the nodes 101 and edges 102 included in the road information can be represented by latitude, longitude, and altitude, or can be represented by an orthogonal coordinate system with the latitude, longitude, and altitude of an arbitrary location as the reference. In the case of using an orthogonal coordinate system, for example, the orthogonal coordinate system can be set in such a way that the latitude corresponds to the Y-axis direction, the longitude corresponds to the X-axis direction, and the altitude corresponds to the Z-axis direction. In addition, IDs that can be specified are assigned to the nodes 101 and edges 102 respectively.

[0064] The three-dimensional map acquisition unit 22 acquires a three-dimensional map and outputs it to the script generation unit 50. It is assumed that the three-dimensional map acquisition unit 22 acquires the three-dimensional map by reading a three-dimensional map pre-stored in a storage medium or the like, but it is not limited thereto. Figure 3 This is a diagram illustrating a three-dimensional map. In the three-dimensional map, in addition to the three-dimensional model of the environment in which the autonomous vehicle travels, it may also include Figure 3 the lane area 103, sidewalk area 104, road markings (such as the lane markings 105 in the lane, the markings 106 in the sidewalk area, crosswalks 107, stop lines 108, etc.), road signs 109, traffic lights 110, and other ground object information shown in the figure. In addition, the position information on the three-dimensional map can be represented by latitude, longitude, and altitude in the same way as the position information of the nodes 101 and edges 102 included in the road information, or can be represented by an orthogonal coordinate system with the latitude, longitude, and altitude of an arbitrary location as the reference.

[0065] The ODD condition acquisition unit 23 acquires ODD conditions and outputs them to the script generation unit 50. ODD is an abbreviation for Operational Design Domain. It is assumed that the three-dimensional map acquisition unit 22 acquires the ODD conditions by reading ODD conditions pre-stored in a storage medium or the like, but it is not limited thereto. The ODD conditions are specific conditions related to the designed driving environment on the premise that the autonomous driving system operates normally, and include road conditions, geographical conditions, environmental conditions, etc. The road conditions include the position, type, school area, position of the bus-only lane, marking type, number of lanes and speed limit, presence or absence of lanes, sidewalks, etc. of the infrastructure sensors 111 that can be utilized by autonomous driving. The geographical conditions include imaginary markings for stopping the autonomous vehicle. The environmental conditions include the weather in which the autonomous vehicle travels, environmental conditions indicating whether GNSS can be utilized, etc. In the ODD conditions, vehicle information also includes vehicle dimensions, the performance of sensors mounted on the vehicle, and the installation position and posture relative to the vehicle origin. Here, the infrastructure sensor 111 refers to a camera or LiDAR installed near roadside equipment or the like, and the recognition results can be flexibly utilized in the autonomous vehicle.

[0066] The driving route generation unit 30 calculates the driving route and driving time of the vehicle to the destination based on the destination information and road information. The driving route generation unit 30 outputs driving route information such as the calculated driving route and driving time to the observation area determination unit 40 and the scenario generation unit 50. In addition, the starting point of the driving route can use the current position of the vehicle itself or a set position.

[0067] The observation area determination unit 40 includes an observation area switching position calculation unit 41 and an observation area calculation unit 42. When the driving route intersects with the moving path of other moving objects based on the driving route and road information, the observation area switching position calculation unit 41 calculates the observation area switching position for switching the observation area from the first observation area to the second observation area. For example, the observation area switching position calculation unit 41 calculates the observation area switching position based on the position of the intersection of the driving route and the lane in which other vehicles are traveling, as will be described later. In addition, the observation area calculation unit 42 calculates the second observation area corresponding to the observation area switching position based on the driving route, road information, and observation area switching position. The observation area determination unit 40 outputs the calculated observation area switching position and the second observation area to the scenario generation unit 50. In addition, the first observation area is, for example, the observation area when the vehicle travels straight, and the second observation area is, for example, the observation area when the vehicle turns left or right.

[0068] The scenario generation unit 50 includes a path deviation amount estimation unit 51 and an evaluation scenario generation unit 52. The path deviation amount estimation unit 51 calculates the deviation amount of deviation from the target path when the vehicle travels on the driving route. The evaluation scenario generation unit 52 generates an evaluation scenario for evaluating the blind spot area observed from the vehicle based on the calculated deviation amount, driving route, road information, and ODD conditions. At this time, the evaluation scenario generation unit 52 generates multiple evaluation scenarios for different environmental conditions (such as weather or driving time period) included in the ODD conditions for the driving route. The generated evaluation scenarios are output to the recognition judgment unit 61.

[0069] For each generated evaluation scenario, the recognition judgment unit 61 simulates the sensing of the second observation area based on the sensors mounted on the vehicle itself to obtain sensing data. And based on the sensing data, it judges whether the recognition of other moving objects located in the second observation area is successful. The risk map generation unit 62 integrates the recognition success or failure results of other moving objects in the second observation area for multiple evaluation scenarios to generate a risk map showing the blind spot area in the second observation area overlapping with the second observation area. The generated risk map is output to the external function 70.

[0070] Next, use Figure 4 to illustrate the processing performed by the blind spot area estimation device 1. Figure 4This is a diagram showing a flowchart of the processing performed by the blind spot area estimation device 1. In this embodiment, a simulator is used to estimate the blind spot areas generated when the host vehicle travels from the autonomous driving start position to the destination before autonomous driving.

[0071] In step S101, the road information acquisition unit 21, the three-dimensional map acquisition unit 22, and the ODD condition acquisition unit 23 each read various data from the storage device, specifically, read road information, a three-dimensional map, and ODD conditions. In step S102, the destination acquisition unit 20 acquires destination information. In addition, the order of step S101 and step S102 is not limited to this, and these two steps may also be performed as one step.

[0072] In step S103, the driving path generation unit 30 calculates the driving path from the autonomous driving start position to the destination. The driving path is represented using Figure 2 the nodes and edges included in the road information as shown. Regarding the nodes and edges to be passed through, for example, the length of the edge can be used as a cost (weight) to obtain using Dijkstra's algorithm.

[0073] In step S104, the observation area determination unit 40 calculates the observation area switching position and the second observation area. The second observation area is calculated for each observation area switching position as described later. Use Figure 5 and Figure 6 to explain the details of step S104. Figure 5 This is a diagram showing a flowchart in step S104. Figure 6 This is a diagram showing an example of road information and a driving path. In Figure 6 , the solid line edge 203 connecting node 201 to node 202 and the nodes represent the driving path. In contrast, the dashed line edges 204, 205, and 206 represent road information. Edge 204 is the oncoming path on which oncoming vehicles travel, edge 205 is the parallel path on which vehicles parallel to the host vehicle travel, and edge 206 is another path that is neither an oncoming path nor a parallel path. Figure 6 The driving path of Figure 2 represents a path in which the host vehicle turns right from a one-way two-lane road to a one-way one-lane road at the same T-junction as

[0074] In Figure 5 step S201, the observation area switching position calculation unit 41 extracts the intersection points of the edges included in the road information and the edges of the driving path. If it is Figure 6In this case, two sides included in side 204, i.e., side 207 representing the lane on the central side and side 208 representing the lane on the outer side, are extracted as the sides that cross the driving path of the own vehicle. The extraction of the crossing sides can be performed by detecting the crossing of line segments (sides) on a two-dimensional coordinate. The extracted sides are recorded in pairs with IDs. In addition, in Figure 6 nodes and sides showing the moving path of the vehicle are described, but in the case where nodes and sides showing the moving path of a pedestrian or the moving path of a bicycle are included in the road information, intersections are also extracted in the same manner as described later.

[0075] In Figure 5 step S202, the observation area switching position calculation unit 41 calculates the position of the intersection of the extracted side and the driving path. The position of the intersection is calculated, for example, as the coordinates of the intersection of line segments on a two-dimensional coordinate. At this time, since only two-dimensional information is used, there may be a place where the roads on the elevated road and the ground road do not actually cross but the position of the intersection is calculated. Therefore, based on the height information included in the side, for example, when the height difference between the crossing sides is 2 m or more, processing such as invalidating the intersection is performed.

[0076] In Figure 5 step S203, the observation area switching position calculation unit 41 groups the intersections for which the intersection positions are calculated in step S202. The grouping is performed to group the moving paths of other moving bodies that should be recognized as a single entity when the own vehicle turns right or left along the driving path. Specifically, the grouping is a process of grouping intersections in which the category of the moving path shown by the side corresponding to the intersection and the road where the moving path is located are the same. As the category of the moving path, for example, the moving path of a vehicle, the moving path of a pedestrian, the moving path of a bicycle, etc. can be cited. If it is Figure 6 the example, since side 207 corresponding to intersection 209 and side 208 corresponding to intersection 210 are the same in that the category of the moving path is the moving path of an automobile and the road where the path is located is also the same, intersection 209 and intersection 210 are grouped into the same group.

[0077] In Figure 5 step S204, the observation area switching position calculation unit 41 calculates the observation area switching position for each group of intersections based on the grouping result of step S203. For example, the observation area switching position calculation unit 41 calculates the position obtained by retracing the driving path from the position of the intersection that is assumed to be the first to be passed by the own vehicle among the intersections included in the same group by a predetermined distance (offset) as the observation area switching position. In Figure 6In the case where, among the intersection points 209 and 210 included in the same group, the intersection point 210 is the one where it is assumed that the own vehicle will pass first. Therefore, the observation area switching position calculation unit 41 calculates the position 211 obtained by retracing the driving path from the intersection point 210 by a preset distance as the observation area switching position.

[0078] In addition, the method for calculating the observation area switching position is not limited to this. For example, as another method, a method of calculating the observation area switching position based on the boundary of the lane can also be used. The observation area switching position calculation unit 41, for example, when the own vehicle turns right as shown in Figure 6 , calculates the position of the intersection point 213 of the boundary 212 between the own lane and the oncoming lane with respect to the side 203 representing the driving path as the observation area switching position. The advantage of the method using the offset is that the observation area switching position can be calculated even in a place where the boundary between the own lane and the oncoming lane is not clear. In contrast, the advantage of using the boundary of the lane is that since the position of the road edge of the oncoming lane can be made the observation area switching position, the second observation area with the observation area switching position set as the starting point will be further optimal as described later. Hereinafter, in step S204, it is assumed that the offset is used to calculate the observation area switching position.

[0079] In Figure 5 step S205, the observation area calculation unit 42 calculates the second observation area for each observation area switching position calculated in step S204. Use Figure 7 , Figure 8 , Figure 9 and Figure 10 to explain step S205 in various situations (left and right turns at intersections, left and right turns at T-intersections without traffic lights) in detail. In addition, in Figures 7 - 10 , the sides and nodes representing the driving path and road information are omitted, and each path is represented by a single line.

[0080] Figure 7 is a diagram illustrating the second observation area when turning right at an intersection, showing a scene where the own vehicle 220 turns right at the intersection along the driving path 221. Here, the dashed line 222 represents the moving path of other vehicles traveling in the oncoming lane, which intersects the driving path 221 of the own vehicle 220 at the intersection points 224 and 225. In addition, the dashed line 223 represents the moving path of pedestrians, which intersects the driving path 221 of the own vehicle 220 at the intersection point 226. In Figure 7In this case, in step S203, the intersection points 224 and 225 are grouped into one group, and the intersection point 226 is grouped into another group. And in step S204, the observation area switching position 227 is calculated for the group of the intersection points 224 and 225, and the observation area switching position 228 is calculated for the group of the intersection point 226. Based on such a premise, in step S205, the second observation area 229 corresponding to the observation area switching position 227 and the second observation area 230 corresponding to the observation area switching position 228 are calculated. In addition, in Figure 7 it is described that the direction parallel to the moving path 222 of the other vehicle is set as the X-axis direction, and the direction perpendicular to the moving path 222 of the other vehicle is set as the Y-axis direction. In addition, the second observation areas 229 and 230 are rectangular shapes, the width in the Y-axis direction is set as the longitudinal width Lv, and the width in the X-axis direction is set as the lateral width L h for description.

[0081] The second observation area 229 is set in such a way that the entire width of the oncoming lane is included in the second observation area 229, and is set in such a way that it includes the range on the positive X-axis side starting from the position of the intersection point 224, which is the closest to the front side on the driving path of the own vehicle, among the intersection points (224, 225) of the group corresponding to the observation area switching position 227. The longitudinal width Lv of the second observation area 229 is set, for example, as the sum of the widths of the oncoming lanes where the moving path 222 used in the extraction of the intersection points 224 and 225 is located. The lateral width L of the second observation area 229 h is obtained by the arithmetic expression (1).

[0082]

Equation 1

[0083] L h = v max * t motion + v max * t sensor + v max * t recog (1)

[0084] Here, L h is the lateral width [m], v max is the upper limit speed [m / s] of the lane intersecting the driving path of the own vehicle, t motion is the time [s] from when passing through the observation area switching position 227 to when passing through the second observation area 229, t sensor is the acquisition period [s] of the sensor data used for observation, t recog is the processing time [s] for object recognition.

[0085] Next, the second observation area 230 is described. The second observation area 230 has a lateral width L hSet the center to become the intersection point 226. The horizontal width of the second observation area 230 is obtained by Equation (1). At this time, v max is the speed [m / s] assumed when a pedestrian or a bicycle moves. The vertical width Lv of the second observation area 230 is set in such a way that the observation area switching position 228 is taken as the starting point and the area of the crosswalk is included in the second observation area 230.

[0086] Figure 8 This is a diagram showing the second observation area when turning left at an intersection, and shows a scene where the own vehicle 220 turns left at the intersection along the driving path 221. Here, the dashed line 222 indicates the moving path of another vehicle passing by the own vehicle and intersects the driving path 221 of the own vehicle 220 at the intersection point 232A. In addition, the dashed line 223 indicates the moving path of a pedestrian and intersects the driving path 221 of the own vehicle 220 at the intersection point 231A. In Figure 8 this case, at step S203, the intersection point 231A is grouped into one group, and the intersection point 232A is grouped into another group. And at step S204, the observation area switching position 231B is calculated for the intersection point 231A, and the observation area switching position 232B is calculated for the group of the intersection point 232A.

[0087] Based on such a premise, at step S205, the second observation area 233 corresponding to the observation area switching position 231B and the second observation area 234 corresponding to the observation area switching position 232B are calculated. Since the calculation method of the second observation area 233 is the same as that of Figure 7 the observation area 230, the calculation method of the second observation area 234 will be described here. In addition, in Figure 8 this, the direction parallel to the moving path 222 of another vehicle is set as the X-axis direction (the first direction), and the direction perpendicular to the moving path 222 of another vehicle is set as the Y-axis direction for explanation. The second observation area 234 is a rectangular shape extending in the negative X-axis direction from the position of the intersection point 232A. Regarding the second observation area 234, the vertical width Lv (the width in the Y-axis direction) of the second observation area 234 is set to the width from the position on the left side of the own vehicle to the road end. The position on the left side of the own vehicle is calculated based on the position of the own vehicle and the width of the own vehicle. The horizontal width L h (the width in the X-axis direction) of the second observation area 234 is obtained by Equation (1).

[0088] Figure 9This is a diagram showing the second observation area when turning right at a T-junction without traffic lights, depicting a scenario where the vehicle 220 turns right at a T-junction without traffic lights along the driving path 221. Here, the dashed line 222A represents the movement path of other vehicles on the lane through which the vehicle turns right, intersecting the driving path 221 of the vehicle 220 at the intersection point 237. The dashed line 222B represents the movement path of other vehicles on the lane into which the vehicle turns right through the right turn, intersecting the driving path 221 of the vehicle 220 at the intersection point 238. The dashed line 223 represents the movement path of pedestrians, intersecting the driving path 221 of the vehicle 220 at the intersection point 236. In Figure 9 this case, in step S203, since the roads where the movement paths are located and the types of the movement paths are the same for the intersection points 237 and 238, they are grouped into one group, and the intersection point 236 is grouped into another group. And, in step S204, the observation area switching position 240 is calculated for the group of the intersection points 237 and 238, and the observation area switching position 239 is calculated for the group of the intersection point 236. Based on such a premise, in step S205, the second observation areas 242 and 243 corresponding to the observation area switching position 240 and the second observation area 241 corresponding to the observation area switching position 239 are calculated. In addition, since the calculation method of the range of the second observation area 241 is the same as that of Figure 7 the second observation area 230, the calculation method of the second observation areas 242 and 243 is described here. In addition, in Figure 9 this, the direction parallel to the movement path 222A of other vehicles is set as the X-axis direction, and the direction perpendicular to the movement path 222A of other vehicles is set as the Y-axis direction for description. Additionally, the second observation areas 242 and 243 are rectangular in shape, with the width in the Y-axis direction set as the vertical width Lv and the width in the X-axis direction set as the horizontal width L h for description.

[0089] When the observation area calculation unit 42 has only intersection points with the same direction of the movement paths used in the extraction of intersection points, such as the intersection points 224 and 225 shown in Figure 7 , for one observation area switching position, one type of second observation area is calculated, such as the second observation area 229 corresponding to the observation area switching position 227 in Figure 7 . In contrast, when the intersection points in the same group of intersection points have different directions of the movement paths used in the extraction of intersection points, such as the intersection points 237 and 238 shown in Figure 9 , for one observation area switching position, the second observation areas are calculated according to the number of types of the directions of the movement paths, such as the second observation areas 242 and 243 corresponding to the observation area switching position 240 in Figure 9 .

[0090] The second observation area 242 is calculated to have a horizontal width L from the intersection point 237 h (not shown) The range obtained by tracing back the movement path 222A used in the extraction of the intersection point 237. The horizontal width L of the second observation area 242 h is obtained by Equation (1). The vertical width Lv of the second observation area 242 is set to the width of the lane shown by the movement path 222A used in the extraction of the intersection point 237

[0091] The second observation area 243 is calculated to have a horizontal width L from the intersection point 238 h (not shown) The range obtained by tracing back the movement path 222B used in the extraction of the intersection point 238. The horizontal width L of the second observation area 243 h is obtained by Equation (1). The vertical width Lv of the second observation area 243 is set to the width of the lane shown by the movement path 222B used in the extraction of the intersection point 238. Since the second observation areas 242 and 243 correspond to the observation area switching position 240, when the host vehicle reaches the observation area switching position 240, the second observation areas 242 and 243 are switched to. The second observation areas 242 and 243 are realized by a plurality of sensors, 360-degree LiDAR, etc

[0092] Figure 10 It is a diagram showing the second observation area when turning left at a T-junction without traffic lights, showing a scenario where the host vehicle 220 turns left at a T-junction without traffic lights along the driving path 221. Here, the dashed line 222 represents the movement path of other vehicles on the lane that the host vehicle enters through the left turn, and intersects with the driving path 221 of the host vehicle 220 at the intersection point 245. The dashed line 223 represents the movement path of pedestrians, and intersects with the driving path 221 of the host vehicle 220 at the intersection point 244. In Figure 10 this case, in step S203, the intersection point 244 is grouped into one group, and the intersection point 245 is grouped into another group. And, in step S204, the observation area switching position 246 is calculated for the group of the intersection point 244, and the observation area switching position 247 is calculated for the group of the intersection point 245. Based on such a premise, in step S205, the second observation area 248 corresponding to the observation area switching position 246 and the second observation area 249 corresponding to the observation area switching position 247 are calculated. In addition, since the calculation method of the second observation area 248 is the same as that of Figure 7 the second observation area 230, the calculation method of the second observation area 249 is described here

[0093] The second observation area 249 is calculated to have a horizontal width L from the intersection point 245 h The range obtained by tracing back the movement path 222 used in the extraction of the intersection point 245. The horizontal width L of the second observation area 249h It is obtained by formula (1). The longitudinal width Lv of the second observation area 249 is set as the width of the lane shown in the moving path 222 used in the extraction of the intersection 245.

[0094] The longitudinal width of the second observation area 249 is set as the width of the oncoming lane corresponding to the observation area switching position 247 and used when calculating the intersection 245 with the driving path of the host vehicle. Next, the calculation method of the lateral width of the second observation area 249 will be described. Regarding the lateral width of the second observation area 249, starting from the position 245 of the intersection, the value obtained by (1) is used for the rear in the traveling direction of the oncoming lane.

[0095] Record the second observation area calculated in step S104 together with the observation area switching position. The dead angle area estimation device 1 repeats Figure 5 After step S205, end Figure 4 Step S104 of Figure 4 and enter

[0096] In Figure 4 In step S105, the script generation unit 50 generates an evaluation script. Use Figure 11 The details of step S105 will be described. Figure 11 It is a diagram showing the flowchart in step S105. Step S105 has steps S401 to S403.

[0097] In step S401, the path deviation amount estimation unit 51 calculates the deviation amount and azimuth deviation amount of the host vehicle relative to the driving path when the autonomous vehicle travels on the set driving path. In this embodiment, it is assumed that the autonomous driving of the vehicle in the target environment is reproduced on the autonomous driving simulator, and based on the position data of the vehicle obtained at this time, the deviation amount and azimuth deviation amount are calculated. However, it can also be based on the position data obtained when the vehicle actually travels along the driving path. The deviation amount and azimuth deviation amount are calculated for example for each side constituting the driving path. In addition, the deviation amount is a quantity indicating how much the vehicle in autonomous driving has deviated from the driving path. The path deviation amount estimation unit 51 calculates, for example, the vertical distance between the side and the position of the vehicle at the time when the assumed vehicle passes through the side as the deviation amount. In addition, the azimuth deviation amount refers to the deviation amount between the orientation of the side and the orientation of the vehicle at the time when the assumed vehicle passes through the side. The path deviation amount estimation unit 51 obtains, for example, the orientation of the side and the orientation of the vehicle as angles, and calculates the absolute value of the difference between these angles as the azimuth deviation amount. Calculate the variance of the deviation amount and azimuth deviation amount. The variance value calculated here is the position error and azimuth error of the assumed host vehicle when traveling on each side of the driving path of the autonomous vehicle.

[0098] After step S401, steps S402 and S403 are repeated according to the number of weather conditions, which is one of the environmental conditions, described in the ODD. In step S402, the evaluation script generation unit 52 generates an evaluation script. The evaluation script is a script for estimating the blind spot area observed from the autonomous vehicle and generating a risk map. The evaluation script includes information on the driving path of the autonomous vehicle, the assumed error during driving, the weather, the sensors mounted on the autonomous vehicle and their installation positions and postures, the second observation area associated with the driving path of the autonomous vehicle, and the observation area switching position. Here, the weather includes weather conditions such as sunny, rainy, cloudy, snowy, and foggy that can affect the sensors. In step S403, the evaluation script generated in step S402 is saved in the storage device. Thus, an evaluation script is generated for each weather condition. In addition, the evaluation script is generated for each weather condition because the observation range of the sensors and the object recognition rate in the recognition function vary significantly depending on the weather. After the dead angle area estimation device 1 repeats steps S402 and S403 according to the number of weather conditions Figure 11 step S402 and S403 of, it ends Figure 4 step S105 of, and proceeds to Figure 4 step S106 of.

[0099] In step S106, the recognition determination unit 61 simulates the sensing of the second observation area based on the sensors mounted on the vehicle for each evaluation script, and determines whether the recognition of other moving objects located in the second observation area is successful. And the risk map generation unit 62 integrates the results of whether the recognition of other moving objects in the second observation area for multiple evaluation scripts is successful, and generates a risk map showing the blind spot area in the second observation area overlapping the second observation area. Use Figure 12 to explain the details of step S106. Figure 12 is a diagram showing the flowchart in step S106. Step S106 includes steps S501 to S505. The dead angle area estimation device 1 performs steps S501 to S505 for each evaluation script according to the number of observation area switching positions included in the evaluation script. In this embodiment, it is assumed that the evaluation script generated for the same driving path of turning right at an intersection as Figure 7 is used to explain step S106. In addition, it is assumed that the driving path of turning right at the intersection illustrated here includes two observation area switching positions, and the Figure 13A and Figure 14A showing the driving path for each observation area switching position are used to explain steps S501 to S505 performed for each observation area switching position.

[0100] Figure 13A is a diagram showing the first observation area switching position and the second observation area when turning right at an intersection. In step S501, asFigure 13A As shown, the recognition and judgment unit 61 moves the own vehicle 500 to the observation area switching position 502. At this time, when it is determined that the vehicle has reached the observation area switching position 502, the observation area is switched to the second observation area 504.

[0101] After step S501, steps S502 to S504 are repeatedly performed to estimate the blind spot area in the second observation area 504 and generate a risk map. The recognition and judgment unit 61 performs steps S502 and S503, and the risk map generation unit 62 performs step S504. Figure 13B It represents the relationship with Figure 13A FIG. is an example of a risk map corresponding to the second observation area 504 shown. As Figure 13B shown, the second observation area 504 in the risk map is represented by a plurality of grids, and each grid is represented as one of a visible area 509, a blind spot area 510, and a conditional blind spot area 511 described later. The conditional blind spot area 511 represents an area that becomes a blind spot area under certain conditions, such as a visible area on a sunny day and a blind spot area on a rainy day, or a blind spot area depending on the driving time.

[0102] To generate Figure 13B a risk map as shown, sensor detection is performed on other vehicles located at positions corresponding to each grid, so sensor data for each grid is required. Therefore, in step S502, other vehicles 507 are virtually arranged at the position of a certain grid in the second observation area 504. Then, virtually, the sensors mounted on the own vehicle are used to detect other vehicles 507, and after obtaining the sensor data, the process proceeds to step S503. In addition, it is assumed that the sensor data reflects the weather set in the evaluation script. For example, in the case of rain or thick fog, the observation area of the LiDAR is restricted, and on a sunny day, the sun position is calculated based on the driving time set in the script and the whitening of the simulated image is performed.

[0103] In step S503, the recognition determination unit 61 uses the sensor data generated in step S502 to determine whether the recognition of the moving object based on the sensor is successful or not. The recognition determination unit 61 inputs the sensor data into the recognition algorithm assumed for the autonomous vehicle. When a moving object of the correct category (e.g., a vehicle) is recognized at the position of the target grid, it is determined that the recognition is successful for the target grid, and in other cases, it is determined that the recognition fails. Further, in the case where a plurality of sensors are mounted, when a moving object of the correct category is recognized at the position of the target grid by using at least one of the sensors, it is determined that the recognition is successful for the target grid. Here, the success or failure of recognition, the sensor data, and the generation conditions of the sensor data are recorded in association with each other. The generation conditions of the sensor data refer to the position and posture of the moving object to be recognized, the category of the moving object, the weather, the time, and the like. After determining the success or failure of the recognition of the moving object in the target grid, the process proceeds to step S504.

[0104] In step S504, the risk map generation unit 62 updates the area category of the target grid in accordance with the following Rules 1 to 8. Here, the number of verification times and the number of successful recognition times recorded at the time of update are used to calculate the recognition success rate of the target grid (number of successful recognition times / number of verification times).

[0105] (Rule 1)

[0106] Regarding the target grid, when the area category before update is the initial value (not determined) and the latest determination is successful recognition, the area category of the target grid is set to the visible area. And the number of verification times and the number of successful recognition times of the target grid are each incremented by 1.

[0107] (Rule 2)

[0108] Regarding the target grid, when the area category before update is the initial value and the latest determination is failed recognition, the area category of the target grid is set to the dead zone. And the number of verification times of the target grid is incremented by 1.

[0109] (Rule 3)

[0110] Regarding the target grid, when the area category before update is the visible area and the latest determination is successful recognition, the area category of the target grid remains set to the visible area. And the number of verification times and the number of successful recognition times of the target grid are each incremented by 1.

[0111] (Rule 4)

[0112] Regarding the object grid, when the area category before update is the visible area and the latest judgment results in recognition failure, set the area category of the object grid as the conditional blind spot area. At this time, record the weather, the category of the recognized object, and the time at the time of recognition failure as the conditions for the conditional blind spot area. Also, increment the verification count of the object grid by 1.

[0113] (Rule 5)

[0114] Regarding the object grid, when the area category before update is the blind spot area and the latest judgment results in recognition success, set the area category of the object grid as the conditional blind spot area. At this time, record the weather, the category of the recognized object, and the time at the time of recognition failure as the conditions for the conditional blind spot area. Also, increment the recognition success count and the verification count of the object grid by 1 respectively.

[0115] (Rule 6)

[0116] Regarding the object grid, when the judgment result before update is the blind spot area and the latest judgment results in recognition failure, still set the area category of the object grid as the blind spot area. At this time, record the weather, the category of the recognized object, and the time at the time of recognition failure as the conditions for the conditional blind spot area. Also, increment the verification count of the object grid by 1.

[0117] (Rule 7)

[0118] Regarding the object grid, when the judgment result before update is the conditional blind spot area and the latest judgment results in recognition success, still set the area category of the object grid as the conditional blind spot area. Also, increment the verification count and the recognition success count of the object grid by 1 respectively.

[0119] (Rule 8)

[0120] Regarding the object grid, when the judgment result before update is the conditional blind spot area and the latest judgment results in recognition failure, still set the area category of the object grid as the conditional blind spot area. At this time, record the weather, the category of the recognized object, and the time at the time of recognition failure as the conditions for the conditional blind spot area. Also, increment the verification count of the object grid by 1.

[0121] Repeat the above steps S502 to S504 for the number of grids in the second observation area 504, and perform sensor data acquisition (step S502), determination of recognition success or failure (step S503), and update of the risk map (step S504) for all grids in the second observation area 504. Regarding Figure 13B For each grid of the second observation area 504 shown, if the initial value is set before update, then for example, in accordance with the above rules 1 and 2, as Figure 13Ba grid determined to be a visible area 509 or a blind spot area 510 as shown. In addition, an area 508 where there are buildings or the like and no moving object can be represented on the risk map as shown Figure 13B by.

[0122] Figure 14A FIG. is a diagram showing a second observation area switching position 513 and a second observation area 514 when turning right at an intersection. Figure 14B It is a diagram showing Figure 14A an example of a risk map corresponding to the second observation area 514 shown. For the second observation area 514 shown Figure 14A as well, sensor data acquisition (step S502), determination of recognition success or failure (step S503), and update of the risk map (step S504) are performed as described above. However, in step S502, a pedestrian 515 is configured as the moving object in the target grid. In addition, a model different from the model for recognizing a car, for example, a model specific to pedestrian recognition, can also be used in the recognition algorithm for inputting sensor data in step S503.

[0123] Regarding Figure 14B each grid of the second observation area 514 shown, if the initial value is set before update, then in step S504, for example, in accordance with the above-mentioned rule 1, as shown Figure 14B by, it is determined to be a visible area 509. The risk map is generated for each driving path. In the case where the driving path has multiple observation area switching positions, as shown Figure 14B by, a plurality of second observation areas are shown on the risk map.

[0124] In addition, when steps S502 to S504 are repeatedly performed, the recognition determination unit 61 assigns the position error and azimuth error of the own vehicle calculated each time in step S401 to the position and orientation of the own vehicle as errors following a normal distribution, for example. And the recognition determination unit 61 determines the success or failure of sensor recognition based on the sensed data observed by the sensor of the own vehicle at the position and orientation to which the position error and azimuth error are assigned, whereby it is possible to calculate the blind spot area observed from the own vehicle in a state where the assumed error is given.

[0125] In this embodiment, it is described that in step S502, the moving bodies are arranged in simulation in a manner corresponding to the positions of the respective grids in the second observation area 504, and sensor detection is performed on the arranged moving bodies, but it is not limited thereto. For example, as step S502, in simulation, another vehicle 507 travels at the upper limit speed along the moving paths 505 and 506 in the second observation area 504, and sensor data during this period is acquired. At this time, the position of the other vehicle 507 is associated with the sensor data in advance, so that it is known which sensor data corresponds to the position of the other vehicle 507. Also, in step S503, it is possible to extract the sensor data when the other vehicle 507 is at the position of the target grid from the sensor data associated with the position of the other vehicle 507, and based on the extracted sensor data, determine whether the identification of the moving body for the target grid is successful. Additionally, in the case where the moving bodies are pedestrians as shown in Figure 14A , and sensor data is acquired while moving these moving bodies, the moving path (the moving path 512 in Figure 14A ) moves in both directions. The moving speed at this time is a speed suitable for pedestrians.

[0126] After the generation of the risk map for one evaluation script is completed, in step S505, the risk map generated at that time is entered in association with the script.

[0127] It is assumed that Figure 14B the shown risk map is the risk map for the evaluation script when the weather is rainy and turning right at an intersection. Here, further consider the risk map for the evaluation script with the same driving path but different weather, that is, the risk map for the evaluation script when the weather is sunny and turning right at an intersection. Since the risk map is generated for each driving path, the risk map for the evaluation script when the weather is sunny and turning right at an intersection is generated by updating the risk map of Figure 14B with the same driving path. An example of the risk map updated based on the evaluation scripts with the same driving path but different weather is shown in Figure 14C . In the risk map with rainy weather shown in Figure 14B , the grids G1 to G4 are judged as dead - end areas 510. In contrast, in the case of sunny weather, since the observation area of the sensor also expands, in step S503, for grids G1 to G3, it is newly judged that the identification is successful, and in step S504, in accordance with rule 5, the area category of grids G1 to G3 is set as a conditional dead - end area 511.

[0128] Through the above series of processes, the blind spot area observed from the vehicle is calculated and a risk map is generated. In addition, although the method of generating a risk map for each script has been described this time, it is also possible to generate a risk map by overlapping the blind spot areas for each evaluation script traveling on the same path.

[0129] Finally, in step S107, the generated risk map is output to the external function 70.

[0130] In addition, in a case where it is difficult to distinguish the movement path of pedestrians and the movement path of bicycles because a part of the sidewalk becomes a bicycle-only road or the like, a bicycle can be assumed as another moving body instead of a pedestrian.

[0131] In addition, the evaluation script generation unit 52 generates evaluation scripts for each type of weather included in the environmental conditions, but for example, it is also possible to generate evaluation scripts for each driving period of the vehicle.

[0132] (Effect)

[0133] By implementing the above series of processes, it is possible to appropriately set the observation area even in an environment including actions such as turning left and right. And when the vehicle control device controls the vehicle to drive automatically in accordance with the driving path, the observation area switching position, and the observation area calculated by the blind spot area estimation device 1, for the blind spot area displayed on the risk map, for example, it is confirmed visually by a person, and if there is any problem, the automatic driving is stopped, so that it is possible to perform corresponding operations such as switching to normal driving.

[0134] As described above, according to the present invention, it is possible to estimate the blind spot area in the observation area of the vehicle.

[0135] Embodiment 2

[0136] Figure 15 It is a diagram showing the structure of the blind spot area estimation device of Embodiment 2. As Figure 15 shown, the blind spot area estimation device 1 of the present embodiment further includes a sensor information acquisition unit 24 and an infrastructure sensor configuration calculation unit 71. Although the external function 70 of Embodiment 1 is not included in the blind spot area estimation device 1, the infrastructure sensor configuration calculation unit 71 of the present embodiment is included in the blind spot area estimation device 1. In addition, since the structure of the blind spot area estimation device 1 is the same as that of Embodiment 1 except for the sensor information acquisition unit 24 and the infrastructure sensor configuration calculation unit 71, the description of the parts common to Embodiment 1 is omitted.

[0137] As mentioned above, infrastructure sensors refer to sensors such as cameras and LiDAR installed in environments such as near traffic lights or roadside equipment, and are installed near intersections and in places where blind spots are likely to form. The autonomous driving car can supplement the recognition of blind spots observed from the vehicle by using the sensor data obtained by the infrastructure sensors. The infrastructure sensor configuration calculation unit 71 uses the risk map obtained in Example 1 to calculate the configuration of the infrastructure sensors in a way that minimizes the blind spots of the vehicle.

[0138] The sensor information acquisition unit 24 acquires the sensor information 14 of the infrastructure sensor. Figure 16 shows an example of sensor information of infrastructure sensors. Sensor information 14 is information related to infrastructure sensors that are predetermined to be installed in the driving environment, and includes parameters for simulating information that can be obtained by various infrastructure sensors and sensor prices. As parameters for simulating information that can be obtained by various infrastructure sensors, if it is LiDAR, it has the number of layers, observation range, observation longitudinal angle, data acquisition cycle, etc., and if it is a camera, it has image size, pixel size, lens model, etc.

[0139] Figure 17 is a diagram showing a flowchart of the processing performed by the infrastructure sensor configuration calculation unit 71. As step S601, the infrastructure sensor configuration calculation unit 71 estimates the candidate of the installation position of the infrastructure sensor. In this embodiment, the position of the road sign included in the three-dimensional map is used as the candidate of the installation position of the infrastructure sensor. In this case, if it is a LiDAR, it is set to be set at a position 1.5m above the ground, and if it is a camera, it is set to be set from the top of the road sign to the oblique downward direction.

[0140] Next, as step S602, the infrastructure sensor configuration calculation unit 71 estimates the blind spot area of ​​the vehicle displayed on the risk map, which can be observed by the infrastructure sensor when the infrastructure sensor is installed. The estimation of the observable blind spot area is performed as follows: Figure 12 As in step S503 in the flowchart of , a model of a moving body such as a vehicle or pedestrian is set in the blind spot area where it is desired to verify whether it can be observed, and sensor data that can be obtained by the predetermined sensor is generated and input into the recognition algorithm using the generated sensor data. At this time, if the blind spot area to be verified whether it can be observed is set as a conditional blind spot area such as rainy days or time (nighttime), the matching conditions are combined to generate sensor data. If the blind spot area to be verified whether it can be observed is a normal blind spot area, it is set to generate sensor data according to the weather conditions and time recorded in the ODD conditions.​​​

[0141] The process of step S602 is repeated for all candidates of the installation positions of the infrastructure sensors, respectively corresponding to the sensor types input as sensor information 14, and further repeated corresponding to a plurality of postures. Thus, for all candidates of the installation positions of the infrastructure sensors, candidates for a plurality of installation postures of the sensors with predetermined settings are generated respectively. As candidates for a plurality of postures, as long as the LiDAR is generated every 2° in the pitch direction within the range of ±10°, the camera is generated every 2° in the pitch direction within the range of ±10° and every 90° in the yaw direction within the range of ±180°, etc. The set of the installation position, installation posture, sensor type, and dead angle area that can be observed of the infrastructure sensors used in this estimation is stored to optimize the configuration of the infrastructure sensors. In addition, when the information of the infrastructure sensors already installed in the environment and the dead angle area that can be observed in the generated risk map are available in the ODD conditions, they are used in the estimation of the configuration of the infrastructure sensors in step S603.

[0142] Next, as step S603, the infrastructure sensor configuration calculation unit 71 estimates the combination of the infrastructure sensors, installation positions, and installation postures set by combinatorial optimization, that is, the configuration of the infrastructure sensors. In step S603, using the installation positions, installation postures of the sensors calculated in the process of step S602, and the information of the dead angle areas that can be observed for each sensor type, the sensor configuration that eliminates all dead angle areas within the target area to be eliminated is estimated. In this embodiment, the combination that satisfies the following constraint conditions is calculated using the greedy method as a set covering problem. Since there are known methods for solving combinatorial optimization problems, for details, please refer to such known methods. In addition, when there are existing infrastructure sensors, although optimization is considered in such a way that the existing infrastructure sensors must be installed, the price of the existing infrastructure sensors is not added to the sensor price.

[0143] Constraint conditions

[0144] (Condition 1) All dead angle areas can be observed by a number of sensors equal to or more than the set number of sensors.

[0145] (Condition 2) The total of the prices of the installed sensors is within the pre-set price.

[0146] (Condition 3) Multiple sensors are not configured at the same sensor installation position.

[0147] Here, the set number of sensors refers to the number considering redundancy, which is set because in areas such as school zones where there may be multiple pedestrians, some pedestrians may block others. In addition, in this embodiment, it is set to observe all blind spots of the vehicle using one or more sensors. As a calculation result of the configuration of infrastructure sensors, multiple configurations that meet the above constraints are sometimes calculated. At this time, for example, a group of multiple configurations and the observed range when the sensors are configured, such as the configuration with the smallest total sensor price, the configuration with the fewest number of sensors, and the configuration with the most sensing multiplicities, are output. In addition, there may be a case where the number of candidates for the installation position of infrastructure sensors is small and a combination that can observe all blind spots cannot be calculated. In this case, a group of the infrastructure sensor configuration that can observe the most blind spots and the remaining blind spots is output.

[0148] Figure 18A FIG. is a diagram showing a driving path 602 in which the vehicle 601 turns right at an intersection. Figure 18B is for Figure 18A FIG. is an example of a risk map generated for the driving path 602 shown. In Figure 18B the risk map shown, grids G1 to G3 are conditional areas 610 that become blind spots in rainy weather, and grid G4 is a blind spot area 611. Based on such a risk map, Figure 17 steps S601 to S603 are performed. As a result, among the candidates 603A to 603D for the installation position of infrastructure sensors in Figure 18A , for example, a sensor is installed at the candidate 603B for the installation position. At this time, since the blind spot area is narrow, a configuration result of the infrastructure sensor that sets a cheap camera with a limited observation range facing this area and a sensor price of 150,000 are output.

[0149] (Effect)

[0150] Through the above series of processes, using the generated risk map, the configuration of infrastructure sensors for assisting the environmental recognition of an autonomous driving vehicle can be automatically calculated.

[0151] Embodiment 3

[0152] Figure 19 FIG. is a diagram showing the structure of the driving environment generation device 2 according to Embodiment 3. The driving environment generation device 2 according to Embodiment 3 estimates the blind spot area of the vehicle based on sensor data obtained by sensing the observation area using sensors mounted on the vehicle during the actual autonomous driving of the vehicle. Thus, for example, the user can evaluate whether the visible area of the risk map obtained through simulation can be observed during the actual driving of the vehicle. Regarding the driving environment generation device 2 of this embodiment, the path deviation amount estimation unit 51 is used instead of Figure 1The script generation unit 50 of the blind spot area estimation device 1 shown deletes Figure 1 the recognition and judgment unit 61 of the blind spot area estimation device 1 shown. And, the driving environment generation device 2 becomes relative to Figure 1 the structure in which the blind spot area estimation device 1 shown is added with a sensor information analysis unit 25 and a blind spot area drawing unit 72. In this embodiment, the description will be centered on the differences from the blind spot area estimation devices of the first and second embodiments.

[0153] The sensor is mounted on the vehicle and acquires information around the vehicle during driving. The sensor information analysis unit 25 analyzes the information acquired by the sensor. Specifically, the sensor information analysis unit 25 detects a moving body from the data acquired by the external recognition sensor mounted on the vehicle. In addition, when a sensor such as a camera capable of predicting the weather is mounted, the weather is estimated, and when a GNSS receiver is mounted, the GNSS reception state is estimated.

[0154] The path deviation amount estimation unit 51 does not calculate the error (variance) with respect to the sides constituting the driving path in the first and second embodiments, but calculates the values of the absolute position error and the absolute azimuth error with respect to the sides. The absolute position error is set as the absolute value of the perpendicular distance from the position of the vehicle to the side (straight line segment) constituting the driving path. The absolute azimuth error is set as the absolute value of the orientation of the autonomous vehicle in the two-dimensional coordinate and the orientation of the side constituting the driving path in the two-dimensional coordinate.

[0155] The risk map generation unit 62 generates a risk map showing the observation state of the vehicle with respect to the surrounding environment based on the sensing data sensed by the sensors mounted on the vehicle, for example, the information of the moving body detected by the sensor information analysis unit 25. At this time, a known method such as an inverse sensor model is used to generate an occupancy grid map around the vehicle. At this time, the state of each grid is set to four states: observable, unobservable, occupied by a moving body, and occupied by a ground object.

[0156] The blind spot area drawing unit 72 overlaps the generated risk map with the second observation area and then displays it to the user. In Figure 20A drawing example of the blind spot area drawing unit 72 is shown. In the blind spot area drawing unit 72, for example, the position of the own vehicle 700, the driving path 702 of the own vehicle position 700, the observation area switching position 703, the second observation area 704, other vehicles 705 observed by sensors mounted on the own vehicle, etc. are drawn on the bird's-eye view 701 of the surrounding environment and presented to the user. At this time, the second observation area 704 is displayed in a manner where numbers are posted as labels. Moreover, the blind spot area drawing unit 72 draws an occupancy grid map 706 corresponding to the bird's-eye view 701 of the surrounding environment and presents it to the user. In the occupancy grid map 706, each grid is represented as one of a visible area 707 (observable state) that can be observed from the own vehicle, a blind spot area 708 (unobservable state) observed from the own vehicle, a moving body area 709 where a moving body exists (moving body occupancy state), and a ground object area 710 occupied by ground objects such as buildings (ground object occupancy state). Additionally, at the same time, the blind spot area drawing unit 72 draws the date and time, weather, the size of the second observation area, the category and position of the detected moving body, the state of the sensors mounted on the own vehicle, and the position error of the own vehicle relative to the target path as a display showing the driving conditions and states. Thereby, after confirming the position of the own vehicle, the second observation area, the blind spots observed from the own vehicle, the positions of other moving bodies, the state of the sensors, etc., the operation of the system can be verified.

[0157] (Effect)

[0158] By implementing the above series of processes, it is possible to verify whether the required area can be observed while the own vehicle is driving.

[0159] As described above, although the embodiments of the present invention have been described, the present invention is not limited to the above embodiments, and various changes can be made without departing from the scope described in the claims. For example, the above embodiments have described the present invention in detail, and it is not necessarily required to have all the structures described. Additionally, the structures of other embodiments can be added to the structure. In addition to this, additions, deletions, and replacements can also be made to a part of the structure.

Claims

1. A blind spot area estimation device for estimating a blind spot area in an observation area of ​​a vehicle that sets an observation area and performs automatic driving based on data acquired by sensing the observation area with a sensor, wherein the blind spot area estimation device is characterized by comprising: a travel route generating unit that calculates a travel route to a destination of the vehicle; an observation area switching position calculation unit for calculating an observation area switching position for switching the observation area from the first observation area to the second observation area when the travel path intersects a travel path of another moving body; an observation region calculation unit configured to calculate the second observation region corresponding to the observation region switching position; a scenario generating unit for generating a plurality of evaluation scenarios having different predetermined environmental conditions for the driving route; an identification judgment unit, which simulates, for each of the evaluation scenarios, sensing of the second observation area by the sensor mounted on the vehicle under relevant environmental conditions, and judges whether the identification of the other moving body located in the second observation area is successful or not; and A risk map generating unit integrates the recognition success or failure results of the other moving objects in the second observation area for a plurality of the evaluation scenarios to generate a risk map showing a blind spot area in the second observation area superimposed on the second observation area.

2. The blind spot area estimation device according to claim 1, characterized in that: The observation area switching position calculation unit calculates, as the observation area switching position, a position obtained by tracing back the travel route by a predetermined distance from the position of the intersection point between the travel route and the movement route.

3. The blind spot area estimation device according to claim 1, characterized in that: The observation area calculation unit calculates the position of the second observation area with respect to the first direction based on the position of the intersection of the driving path and the moving path, when a direction parallel to the moving path is set as the first direction, and calculates the width of the second observation area with respect to the first direction based on the moving speed of the other moving body, the time from when the other moving body passes through the observation area switching position to when the other moving body passes through the second observation area, the sensor data acquisition cycle of the sensor, and the processing time of the object recognition.

4. The blind spot area estimation device according to claim 1, characterized in that: When the travel route intersects with the plurality of movement routes respectively and there are a plurality of intersection points between the travel route and the movement routes, the observation area switching position calculation unit divides the plurality of intersection points into groups and calculates the observation area switching position for each group.

5. The blind spot area estimation device according to claim 4, characterized in that: The movement paths include at least two types: a movement path of a vehicle and a movement path of a pedestrian. The observation area switching position calculation unit classifies the plurality of intersection points into the groups according to the type of the movement route intersecting the travel route.

6. The blind spot area estimation device according to claim 4, characterized in that: The observation area switching position calculation unit calculates, as the observation area switching position, a position obtained by tracing back the travel route by a predetermined distance from a position of the intersection that the vehicle is expected to pass first among the intersections in the group.

7. The blind spot area estimation device according to claim 1, characterized in that: The method further comprises a route deviation estimation unit for calculating a position error and an orientation error of the vehicle relative to the travel route, The recognition and determination unit simulates sensing of the second observation area by the sensor mounted on the vehicle at the position and orientation to which the position error and the orientation error are given.

8. The blind spot area estimation device according to claim 1, characterized in that: The risk map generation unit represents the second observation area with a plurality of grids in the risk map, and classifies the area type of each of the plurality of grids into a visible area, a blind spot area, or a conditional blind spot area.

9. The blind spot area estimation device according to claim 8, characterized in that: The visible area represents an area determined by the recognition judgment unit as a successful recognition in a plurality of the evaluation scenarios. The blind spot area represents an area where the recognition judgment unit judges that the recognition fails in a plurality of the evaluation scenarios. The conditional blind spot area indicates an area determined by the recognition judgment unit as a successful recognition in a part of the plurality of evaluation scenarios.

10. The blind spot area estimation device according to claim 1, characterized in that: The environmental condition is weather or driving time period.

11. The blind spot area estimation device according to claim 1, characterized in that: The system further includes an infrastructure sensor configuration calculation unit that calculates the configuration of the infrastructure sensors based on the risk map so that the infrastructure sensors configured in the environment can sense the blind spot area.

12. A vehicle control device, characterized in that: The vehicle is controlled according to the travel path, the observation area switching position, and the observation area calculated by the blind spot area estimating device described in any one of claims 1 to 10.

13. A vehicle, characterized in that: A vehicle control device according to claim 12.

14. A driving environment generation device for estimating a blind spot area in an observation area of ​​a vehicle that sets an observation area and performs automatic driving based on data acquired by sensing the observation area with a sensor, the driving environment generation device being characterized by: a travel route generating unit that calculates a travel route to a destination of the vehicle; an observation area switching position calculation unit for calculating an observation area switching position for switching the observation area from the first observation area to the second observation area when the travel path intersects a travel path of another moving body; an observation region calculation unit configured to calculate the second observation region corresponding to the observation region switching position; a risk map generating unit for generating a risk map showing a state of observation of a surrounding environment by the vehicle based on sensing data obtained by sensors mounted on the vehicle; and The blind spot area drawing unit displays the risk map and the second observation area in an overlapping manner.

15. The driving environment generating device according to claim 14, characterized in that: The blind spot area drawing unit displays weather information estimated based on the sensing data.

Images