Parking assistance device

Abstract

To provide a parking support device that can improve convenience of leaning-type automatic parking.SOLUTION: The parking support device comprises map generating means and running control means. The map generating means registers map including a route extending from a parking start position to a parking position, during running for learning. The running control means makes a vehicle autonomously run on the basis of the map, during automatic parking. The map includes a plurality of routes extending from one parking start position to a plurality of parking positions. Further, the parking support device comprises route selecting means. The route selecting means selects the route on which the vehicle runs out of the plurality of routes, during automatic parking.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present disclosure relates to a parking assistance device.

Background Art

[0002] Patent Document 1 discloses a parking assistance device that guides a vehicle to start learning driving from different starting positions when the parking positions are different when repeating learning driving at a close position in automatic parking with learning driving (hereinafter also referred to as learning-type automatic parking).

[0003] Conventionally, a map used for automatic parking was managed with the starting position as an index. For this reason, in the automatic parking technology, when repeating learning driving at the same starting position, it was regarded as redoing the learning driving, and there was a problem that even if the parking positions in the learning driving were different, only the parking position at the time of the last learning driving was registered.

[0004] The technology of Patent Document 1 is a technology for solving the above problems. According to the technology disclosed in Patent Document 1, when the parking positions are different, by changing the starting position and performing learning driving, a plurality of maps with different starting positions are generated. Therefore, with the technology of Patent Document 1, different parking positions can be registered in the map.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] Here, FIG. 23 is a diagram showing an example of automatic parking processing by a parking support device using the technology of Patent Document 1. In the example of FIG. 23, using the technology of Patent Document 1, a parking position PA is associated with a start position PZ. When the occupant of the vehicle VA instructs the start of automatic parking at the start position PZ and selects to park at the parking position PA, and after the parking support device starts automatic parking for Vehicle 1, when the vehicle VA reaches the position PY, it shows the case where another vehicle VB is parked at the parking position PA. At this time, since the vehicle VA is performing automatic parking according to the map for parking at the parking position PA, if another vehicle VB is parked at the parking position PA, automatic parking cannot be continued.

[0007] For example, if the parking support device also stores a map for parking at the parking position PB, since the space at the parking position PB is available, it could return to the parking start position (not shown) for parking at the parking position PB and start the automatic parking from the beginning, but this is too troublesome. For example, if it were possible to change the map to be referenced to another map during automatic parking and continue the automatic parking, the conventional parking support device cannot do so. Thus, there is room for improvement in terms of convenience in the technology of learning-based automatic parking.

[0008] An object of the present disclosure is to provide a parking support device that improves the convenience of learning-based automatic parking.

Means for Solving the Problem

[0009] To solve the above problems, the parking support device according to the present disclosure includes a map generation means and a travel control means. The map generation means registers a map including a route from a parking start position to a parking position during learning travel. The travel control means autonomously travels the vehicle based on the map during automatic parking. Further, the map has a plurality of routes from one parking start position to a plurality of parking positions. Furthermore, the parking support device includes a route selection means. The route selection means selects a route for the vehicle to travel from a plurality of routes during automatic parking.

Effect of the Invention

[0010] According to the present disclosure, the convenience of learning-type automatic parking can be improved.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Embodiments for Carrying Out the Invention

[0012] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. Note that each of the embodiments described below shows a specific example of the present disclosure. Therefore, each component shown in the following embodiments, the arrangement position and connection form of each component, and the order of each step and each step are merely examples and do not limit the gist of the present disclosure. In addition, among the components in the following embodiments, the components not described in the independent claims are described as optional components.

[0013] In addition, each figure is a schematic diagram and is not necessarily drawn precisely. In each figure, the same reference numerals are given to substantially the same configurations, and duplicate explanations are omitted or simplified.

[0014] (Learning-based Automatic Parking) FIG. 1 is a diagram showing an example of learning-based automatic parking according to the present embodiment. In learning-based automatic parking, the start position and the parking route are memorized by manually parking in advance. When the occupant activates the automatic parking function at the memorized start position, the vehicle 1 can automatically travel along the parking route when manually parked in advance and park itself.

[0015] More specifically, in learning-based automatic parking, the vehicle 1 generates a map including the position information of ground objects (hereinafter also referred to as ground object information) during manual parking (hereinafter also referred to as learning driving). When automatically parking, the vehicle 1 travels while estimating its position and attitude using the map.

[0016] Here, the ground object refers to an object such as a road surface marking or an obstacle that is around the vehicle and does not move. The position of the ground object may be specified by identifying the azimuth of the object from the positions of the images of the object captured by a plurality of cameras and then specifying the coordinates based on the principle of triangulation. Alternatively, the coordinates may be specified using a distance measuring device such as a sonar or a radar and based on the principle of trilateration from a plurality of positions.

[0017] The map of the learning-based automatic parking includes the position information of ground objects (hereinafter also referred to as ground object information) and the data of the parking route. The parking route is, for example, as shown in FIG. 1, a route connecting the parking start position 15 and the parking position 18. The parking route may be divided at the end points 16 and 17 where the steering angle changes into straight line sections and curve sections. Also, the data of the parking route may be registered as an aggregate of data of a plurality of sections. The data of each section is accompanied by length information, and the data of the curve section may be accompanied by information such as the steering angle and the turning radius. The data of the parking route according to the present embodiment consists of data of a plurality of sections.

[0018] The ground object information may be, for example, data of a plurality of feature points extracted from a camera image. A feature point is a point that can identify the position extracted from the image of the ground object reflected in the camera image, and the data of the feature points registered in the map is accompanied by position information. The position information may be the position information of the ground object specified by the principle of triangulation surveying, or the position information specified by the principle of the above-described triangulation surveying. Regarding the extraction of feature points, since existing methods can be used, detailed description thereof will be omitted.

[0019] During learning driving, the parking support device generates a map including ground object information (data of a plurality of feature points) and a parking route (data of a plurality of sections). During automatic parking, the parking support device estimates the position and attitude of the host vehicle by collating the data of a plurality of feature points read from the map with the data of a plurality of feature points extracted from the camera image. Then, the parking support device controls the vehicle so that the position of the host vehicle sequentially follows a plurality of sections included in the parking route read from the map.

[0020] (Configuration of Vehicle) FIG. 2 is a diagram showing a vehicle 1 to which the parking support device according to the embodiment can be applied. As shown in FIG. 2, the vehicle 1 is equipped with a parking support system 1S. The parking support system 1S includes an operation device 10, an HMI (Human Machine Interface) device 20, a vehicle control device 30, a navigation device 40, a sonar ECU 50, and a parking support ECU 100.

[0021] Note that the vehicle 1 may be further equipped with other devices. Also, in FIG. 2, the operation device 10, the HMI device 20, the vehicle control device 30, the navigation device 40, the sonar ECU 50, and the parking support ECU 100 are illustrated as separate devices, but a part or all of these devices may be integrated.

[0022] The operation device 10, the HMI device 20, the vehicle control device 30, the navigation device 40, the sonar ECU 50, and the parking support ECU 100 will be described later.

[0023] At four locations around the vehicle body of the vehicle 1, cameras 2a, 2b, 2c, and 2d are provided. Hereinafter, when the cameras 2a, 2b, 2c, and 2d are not particularly distinguished, they are simply referred to as the camera 2. Each camera 2 is equipped with a fish-eye lens and has a horizontal field of view of 180 degrees or more (see the dashed line).

[0024] Since each camera 2 is mounted with a downward angle to capture the road surface, when converting the range where the road surface is reflected into the horizontal field of view, the road surface in a range of about 240 degrees is reflected in one camera 2. For example, in the captured images of the side cameras 2a and 2b provided on the left and right sides of the vehicle body of the vehicle 1, the front wheels, the rear wheels, and the side surface of the vehicle body of the vehicle 1 are reflected.

[0025] Note that the installation location and number of the cameras 2 are not limited to the example shown in FIG. 2.

[0026] Also, as shown in FIG. 2, twelve sonar sensors 3a to 3l are installed in the vehicle 1. Hereinafter, when the sonar sensors 3a to 3l are not particularly distinguished, they are simply referred to as the sonar sensor 3. For example, on the left side of the vehicle 1, a sonar sensor 3a is installed on the front side (FLS: Front Left Side) of the vehicle 1, and a sonar sensor 3b is installed on the rear side (BLS: Back Left Side) of the vehicle 1.

[0027] Also, on the right side of the vehicle 1, a sonar sensor 3c is installed on the front side (FRS: Front Right Side) of the vehicle 1, and a sonar sensor 3d is installed on the rear side (BRS: Back Right Side) of the vehicle 1. These four sonar sensors are also called side sonars for detecting obstacles on the side of the vehicle.

[0028] Also, in front of the vehicle 1, sonar sensors 3e (FLC: Front Left Corner), 3f (FL: Front Left), 3g (FR: Front Right), and 3h (FRC: Front Right Corner) are installed in order from the left side in the forward direction of the vehicle 1.

[0029] The sonar sensors 3f and 3g provided on the inner side detect obstacles in the traveling direction when the vehicle 1 travels straight. Also, the sonar sensors 3e and 13h provided on the outer side detect obstacles in the turning direction when the vehicle 1 turns. The sonar sensors 3e and 3h are also called corner sonars. In FIG. 1, the detection ranges of the four sonar sensors 3e, 3f, 3g, and 3h are shown as triangles, but the detection ranges are not limited to the triangular ranges and can detect up to about 10 m from the vehicle. Also, the detection ranges of adjacent sonars are installed so as to overlap each other.

[0030] Also, behind the vehicle 1, sonar sensors 3i (BLC: Back Left Corner), 3j (BL: Back Left), 3k (BR: Back Right), and 3l (BRC: Back Right Corner) are installed in order from the left side in the forward direction of the vehicle 1.

[0031] The sonar sensors 3j and 3k provided on the inner side detect obstacles in the traveling direction when the vehicle 1 reverses. Also, the sonar sensors 3i and 3l provided on the outer side detect obstacles in the turning direction when the vehicle 1 reverses and turns. The sonar sensors 3i and 3l are also called corner sonars. In FIG. 2, the directions detected by the four sonar sensors 3i, 3j, 3k, and 3l and the fan-shaped spreads thereof are illustrated as triangles, but the detection ranges are not limited to the inside of the triangles in the figure and extend. For example, adjacent sonars are installed so that their detection ranges overlap each other. The same applies to the four sonar sensors 3e, 3f, 3g, and 3h on the front side of the vehicle.

[0032] Note that the detection ranges of the sonar sensors 3a, 3b, 3c, and 3d installed on the side of the vehicle 1 are set to be narrower than the detection ranges of the sonar sensors installed in front of and behind the vehicle 1. This is to improve the positional resolution when the parking support ECU 100 detects the parking space on the side of the vehicle 1 by minimizing the overlap of the detection ranges of the side sonars when the vehicle 1 moves.

[0033] Also, each sonar sensor is installed at a height and depression angle that makes it easy to detect surrounding obstacles when the vehicle 1 parks. Note that the installation locations and numbers of the sonar sensors 3a to 3l are not limited to the example shown in FIG. 2.

[0034] Here, in this embodiment, the sonar refers to the sonar system composed of the above-described sonar sensors 3a to 3l and the sonar ECU 50. The sonar ECU 50 is a control device that comprehensively controls the sonar system.

[0035] The sonar sensor 3 emits a directional sound wave and receives the reflected wave. The sonar ECU 50 detects the distance to the obstacle based on the time from when the sonar sensor 3 emits the sound wave until it receives the reflected wave. The sonar ECU 50 detects the vicinity of the vehicle using a number of sonars and identifies the position of the obstacle based on the distances from the plurality of detected sonar sensors 3.

[0036] The sonar system detects an obstacle in the traveling direction of the vehicle 1 using the sonar sensor 3 provided on the bumper of the vehicle 1. When the sonar ECU 50 detects an obstacle, it determines whether the vehicle 1 and the obstacle will collide. When the sonar ECU 50 determines that the vehicle 1 and the obstacle will collide within a predetermined time, it instructs the vehicle control device 30 to activate the automatic brake.

[0037] This ensures that the vehicle 1 does not collide with the front and rear obstacles even during automatic parking. Also, by detecting with the side sonars provided on the left and right side surfaces of the vehicle body of the vehicle 1, it is possible to detect that the side surface of the vehicle body approaches an obstacle due to the inner wheel difference.

[0038] Note that the sonar is not an essential component of the parking assistance system 1S. That is, the vehicle 1 may be configured without a sonar. In this case, for example, the parking assistance ECU 100 may detect the distance to an obstacle by processing the camera image captured by the camera 2. For example, the parking assistance ECU 100 may determine whether the vehicle 1 and an obstacle will collide within a predetermined time. If it is determined that a collision will occur, the parking assistance ECU 100 may instruct the vehicle control device 30 to activate the automatic brake.

[0039] (Hardware Configuration of Parking Assistance ECU) Next, the hardware configuration of the parking assistance ECU 100 will be described. The parking assistance ECU 100 is an example of a parking assistance device. FIG. 3 is a diagram showing an example of the hardware configuration of the parking assistance ECU 100 according to the embodiment. The functions of the parking assistance ECU 100 described later may be implemented in the hardware shown in FIG. 3.

[0040] The parking assistance ECU 100 may include a CPU 101, a ROM 102, a RAM 103, an I / O (input / output interface) 104, an IMP (Image Processor) 105, and a communication I / F (Interface) 106, and may be a computer that connects each element with a bus.

[0041] The parking assistance ECU 100 may house a plurality of elements in one chip. Further, the parking assistance ECU 100 may be configured with a plurality of chips for one element. The bus may not be single, and a combination of a plurality of types of buses may be used.

[0042] For example, the CPU 101, ROM 102, RAM 103, IMP 105, and communication I / F 106 may be housed in one chip and connected with a parallel bus, and the I / O 104 may be configured with a plurality of chips and connected to the chip housing the CPU 101 with a serial bus.

[0043] The parking support ECU 100 executes automatic parking by obtaining information from other devices (e.g., the navigation device 40) or giving instructions to other devices (e.g., the vehicle control device 30) via the communication I / F and the in-vehicle LAN.

[0044] The CPU 101 controls the entire parking support ECU 100. The functions of each part of the parking support ECU 100 may be implemented in the form of programs executed by the CPU 101. The ROM 102 and the RAM 103 correspond to the storage unit, and the ROM 102 corresponds to the non-volatile area. The RAM 103 is used for temporary storage as the working area of the CPU 101. For example, camera images such as display images and detection images, information on detected feature points, etc. are temporarily stored in the RAM 103.

[0045] The IMP 105 is a processor specialized for image processing and parallel processing to improve processing performance. Part of the functions provided by the parking support ECU 100 described later (e.g., the image processing unit 130, the space recognition unit 140, the position estimation unit 150, etc.) may be executed by the IMP 105.

[0046] (Functions of the parking support ECU) Next, the functions of the parking support ECU 100 will be described. FIG. 4 is a block diagram showing an example of the configuration of the parking support system 1S according to the embodiment. As shown in FIG. 4, the cameras 2 (2c(F), 2d(B), 2a(L), 2b(R)) output the captured camera images. The image processing unit 130 of the parking support ECU 100 receives the camera images and performs operations such as generating display images. The display images are output from the notification unit 180 to the HMI device 20.

[0047] The notification unit 180 superimposes a message on the display image or outputs an audio message according to the instruction of the state management unit 110. The notification unit 180 is an example of a notification means, and the HMI device 20 which is the output destination of the message and the state management unit 110 which instructs the output of the message may also be included in the notification means. That is, the notification means is a functional element that conveys information to the occupant. Since the state management unit 110, the notification unit 180, and the HMI device 20 are involved in the notification, they may be referred to as the HMI device 20, etc.

[0048] The state management unit 110 receives the user's operation by the operation device 10 and controls the functions of the parking support ECU 100 according to the user's operation. Here, the touch panel of the navigation device 40 is included in the operation device 10.

[0049] The state management unit 110 receives the position information from the main body (not shown) of the navigation device 40. When performing learning driving, the state management unit 110 adds the position information of the parking start position to the map and records it in the storage unit 170. When performing automatic parking, the state management unit 110 compares the position information of the navigation device 40 with the position information added to the map and selects an available map.

[0050] The image processing unit 130 generates a display image and a detection image. The detection image is, for example, an image in which changes in luminance or color, that is, contrast, are emphasized. The image processing unit 130 extracts feature points from the detection image. The image processing unit 130 extracts, as feature points, points that can be identified as points rather than the surfaces or sides of the image and that correspond to the corners or ends of the lines of the image. The information on the feature points output by the image processing unit 130 includes information on the color and shape of the image and the position information of the feature points on the camera image.

[0051] The space recognition unit 140 identifies the positions and distributions of the feature points. Further, the space recognition unit 140 identifies obstacles, that is, feature points not on the road surface, by analyzing the positions and distributions of the feature points. Then, the space recognition unit 140 recognizes a region without obstacles, that is, a space, based on the positions and distributions of the identified obstacles.

[0052] FIG. 5 is a diagram for explaining an example of a process in which the space recognition unit 140 identifies the positions of feature points. As shown in FIG. 5, the space recognition unit 140 identifies the positions of feature points (subjects) P imaged by the camera 2 of the vehicle 1 based on motion parallax. First, the space recognition unit 140 converts the position of the feature point P on the camera image into the angle (azimuth angle) of the feature point P with respect to the vehicle 1. For example, it is assumed that as the vehicle 1 moves, the camera 2 of the vehicle 1 moves from point A on the Y-axis to point B. Then, the image of the feature point moves on the camera image, and the azimuth angle obtained from the position of the feature point P imaged by the camera 2 changes from θ1 to θ2. Such a change in the azimuth angle (θ1 → θ2) is called motion parallax.

[0053] Since the distance between point A and point B (the length of the line segment AB) is the amount of movement of the vehicle 1, it can be specified from the number of rotations of the wheels. The space recognition unit 140 can specify the XY coordinates (x, y) of the feature point P from the coordinates and length of the line segment AB and θ1 and θ2 based on the principle of triangulation. The position of the feature point P imaged by the camera 2 in the vertical direction within the image corresponds to the height of the feature point P. For example, if the feature point P is the tip of a rod standing perpendicular to the ground, it is imaged above the base of the rod in the screen. Therefore, if the space recognition unit 140 specifies the XY coordinates, it can specify the Z coordinate based on the position in the vertical direction within the image. That is, the space recognition unit 140 can specify the three-dimensional coordinates of the feature point (subject) P by applying the motion parallax appearing in the camera images taken at different times to the principle of triangulation.

[0054] In addition, the space recognition unit 140 recognizes the space based on the parallax distribution. FIG. 6 is a diagram for explaining an example of the process of recognizing the space based on the parallax distribution. FIG. 7 is an example of a graph representing the parallax distribution. For example, as shown in FIG. 6, consider the case where a vertical wall 201 parallel to the traveling route of the vehicle 1 stands up. Let the position of the camera 2 be point A, the intersection of the perpendicular line dropped from point A to the road surface and the road surface (directly below the camera 2) be point B, the intersection of the perpendicular line dropped from point B to the wall 201 and the wall 201 be point C, the intersection of the perpendicular line dropped from point A to the wall 201 and the wall 201 be point D, and consider the case where the road surface continues from below the vehicle 1 to point C. Point E in the figure is a point on the way on the perpendicular line BC. In addition, there are feature points on each of points B, C, D, E, and line segments BC and CD.

[0055] In this case, in FIG. 7, the horizontal axis represents the angle θ indicating the direction of the feature point with 0 degrees directly below the camera 2, and the vertical axis represents the parallax generated at the feature point. As shown in FIG. 7, the graph showing the parallax of the feature point imaged in the direction of the angle θ turns from decreasing to increasing at the angle θc of point C. The reason will be explained below.

[0056] The lower diagram in FIG. 6 shows the parallax caused by the movement of the vehicle at the feature points C and E directly beside the camera 2. The motion parallax of the feature point of point C is θ3, and the motion parallax of the feature point of the intermediate point E is θ4. In this example, the motion parallax θ4 of the nearer point B is larger than the motion parallax θ3 of the farther point C. From this, it can be seen that the closer to the camera 2, the larger the parallax. Therefore, between the points B and C on the road surface (on the line segment BC, in the range of d1 in FIG. 7), the larger the θ, the smaller the parallax. On the other hand, between the points C and D on the wall surface of the wall 201 (on the line segment CD, in the range of d2 in FIG. 7), since the height difference from the camera 2 is smaller, the shorter the distance, and accordingly the smaller the parallax. For this reason, in the example of FIG. 6, between the points C and D on the wall surface of the wall 201, the larger the θ, the larger the parallax. If there were no wall 201, the parallax would continue to decrease as shown by the dotted line in FIG. 7. Therefore, based on the fact that the parallax rises at θc, it can be specified that there is a three-dimensional object continuing above point C.

[0057] In this way, the space recognition unit 140 can identify the position of a three-dimensional object such as the wall 201 by analyzing the distribution of parallax. Also, since there is no discontinuity in the distribution of parallax up to point C (within the range of d1), it indicates that a road surface without steps is visible. Furthermore, since there is no discontinuity in the distribution of parallax between point C and point D (within the range of d2), it is possible to identify the presence of the wall 201. Also, the fact that the wall 201 is vertical can also be identified from the distribution curve. In this manner, the space recognition unit 140 can identify that there is space on the side of the vehicle 1 from the distribution of parallax.

[0058] Also, when approaching the parking position, the space recognition unit 140 detects the parking space and determines whether parking is possible. FIG. 8 is a diagram for explaining an example of a process for determining whether parking is possible. Here, consider the case where, in the situation shown in FIG. 8, the space recognition unit 140 of the vehicle 1A detects the parking space and determines whether parking is possible.

[0059] In the example of FIG. 8, the triangular cone 202 is placed at the back of the parking space. In FIG. 8, although the vehicle 1B is parked in the parking space, assume that the vehicle 1B did not exist during the learning drive. That is, assume that the feature points of the triangular cone 202 at the back and the feature points in the parking space were registered on the map during the learning drive. In that case, the space recognition unit 140 can compare the observed motion parallax with the motion parallax during the learning drive. For example, when imaging the direction of the floor surface of the parking space from the vehicle 1A, the observed motion parallax becomes larger because the distance to the subject is shortened due to the presence of the vehicle 1B, compared to the case where the vehicle 1B does not exist and the floor surface of the parking space is imaged in the camera image.

[0060] For example, when the space recognition unit 140 analyzes the distribution of the motion parallax of the floor surface portion of the parking space and determines that there is a region where the parallax is larger than during the learning drive as described above, it may determine that parking is not possible because there is a three-dimensional object in the parking space.

[0061] Further, when the space recognition unit 140 fails to detect the feature points on the back of the parking space or the floor surface of the parking space that were registered during the learning drive, it may determine that parking is impossible because there is an object blocking them. In the example of FIG. 8, when the space recognition unit 140 fails to detect the feature points representing the triangular cones 202 that were registered during the learning drive, it may determine that parking is impossible.

[0062] In the present embodiment, an example in which the space recognition unit 140 recognizes the space by detection using the camera 2 will be mainly described. However, the space recognition unit 140 may determine whether parking is possible or not by means of obstacle detection means such as a sonar or a radar. For example, when the vehicle 1 is parked, if the distance to the nearest obstacle on the side of the vehicle 1 is equal to or greater than the vehicle width of the vehicle 1, the space recognition unit 140 may estimate that there is a space where parking is possible on the side.

[0063] Further, the space recognition unit 140 may perform processing in combination with detection using the camera 2 and obstacle detection using a sonar or the like. The space recognition unit 140 processes the feature points by detection using the camera 2, and comprehensively determines using the obstacle detection information obtained by the sonar as well. For example, even if the feature points of the triangular cones 202 that were at the back of the parking space during the learning drive do not appear in the camera image during the automatic parking, if the sonar does not detect an obstacle in the range corresponding to the floor surface of the parking space, it may be determined that parking is possible.

[0064] The position estimation unit 150 functions during automatic parking, and estimates the position and attitude of the host vehicle based on the feature points registered in the map and the feature points on the camera image. Further, the map may include a camera image taken at the start of the learning drive.

[0065] Specifically, when starting automatic parking, the position estimation unit 150 collates the information of the feature points detected by the image processing unit 130 (information such as the color and shape of the image and the position information of the feature points on the camera image) with the information of the feature points registered in the map, and identifies the feature points that match the map.

[0066] If the position of Vehicle 1 at the start of automatic parking is close to the position where learning driving started and there is no significant difference in the orientation (posture) of the vehicle body of Vehicle 1, it can be expected that the position of the feature points on the camera image at the start of automatic parking will not be significantly different from the position of the feature points on the camera image at the start of learning driving. Therefore, the position estimation unit 150 may extract the feature points that match the map by comparing the positions of the feature points on the camera images at the start of automatic parking and at the start of learning driving. Alternatively, corresponding feature points may be identified by comprehensive matching processing that adds comparison of image color and shape information to the comparison of the detected arrangement of feature points and the arrangement of feature points registered in the map. Here, the matching process is a process of identifying feature points on the camera image that match the feature points on the map.

[0067] When the position estimation unit 150 identifies the corresponding feature points, it identifies the position and posture (orientation) of the vehicle at the start of automatic parking by comparing the orientation of the feature points registered in the map with reference to the start position of learning driving and the orientation of the feature points on the camera image. For example, since the horizontal position of the feature points on the camera image corresponds to the orientation of the feature points with respect to the vehicle, the position estimation unit 150 can identify the posture of the vehicle based on the horizontal position of the feature points.

[0068] Also, since the map includes the three-dimensional coordinates of the feature points and the vertical position of the feature points on the camera image corresponds to the depression angle, the position estimation unit 150 can identify the distance to the feature points registered in the map based on the vertical position of the feature points. The position estimation unit 150 repeats these processes for the corresponding feature points and identifies the most likely value (maximum likelihood value) as the position and posture of the host vehicle.

[0069] As described above, the position estimation unit 150 identifies the feature points on the map that match the feature points on the image, and estimates the position and posture of Vehicle 1 from the three-dimensional coordinates of the feature points on the map and the position of the feature points on the camera image that match them. These processes are called self-position estimation. The self-position estimation process and the matching process may be processed using existing methods, and detailed descriptions are omitted.

[0070] When performing learning driving, the travel control unit 160 communicates with the vehicle control device 30 regularly to obtain information on the rotation speed of the wheels of the vehicle 1 and the steering angle. Then, based on the obtained information, it calculates the movement amount and the attitude change amount (the change amount of the orientation of the vehicle body of the vehicle 1) of the vehicle 1 per unit time.

[0071] The travel control unit 160 obtains the attitude of the vehicle body of the vehicle 1 by integrating the attitude change amount. The travel control unit 160 obtains the movement vector of the vehicle 1 per unit time from the obtained attitude of the vehicle body and the movement amount of the vehicle 1. The travel control unit 160 obtains the coordinates of the vehicle body at each time by integrating the obtained movement vectors. The travel control unit 160 obtains a travel trajectory graphically represented by a polyline by connecting the coordinates of the vehicle body at each time.

[0072] The above process is called the own vehicle position estimation of the travel control unit 160. Also, the route connecting the own vehicle positions obtained by the own vehicle position estimation is called the parking route by the own vehicle position estimation.

[0073] During learning driving, the travel control unit 160 may send the rotation speed and steering angle of the wheels of the vehicle 1, or the movement amount and movement direction (movement vector) of the vehicle 1 to the map generation unit 120 at each moment as information on the parking route, or may send the estimated own vehicle position to the map generation unit 120 at each moment. Note that the travel control unit 160 of the present embodiment sends the parking route by the own vehicle position estimation to the map generation unit 120 at the end of the learning driving. Since this parking route is a series of movement vectors per unit time, it is graphically represented in a polyline shape when viewed from above.

[0074] When performing automatic parking, the travel control unit 160 estimates the own vehicle position by the above method and at the same time outputs an instruction value to the vehicle control device 30 so as to reproduce the steering angle and the travel distance recorded on the map. The vehicle control device 30 controls the steering angle and the vehicle speed according to the instruction value. However, the steering angle and the vehicle speed follow the change of the instruction value with a delay and may deviate from the instruction value temporarily. Also, the actual steering angle and vehicle speed follow the instruction value with an offset and may deviate constantly. As a result, the route estimated by the travel control unit 160 for the own vehicle position may deviate from the parking route recorded on the map.

[0075] In such a case, the travel control unit 160 performs feedback control in the direction of pulling back the vehicle 1 to the parking route. Specifically, the travel control unit 160 first controls the steering angle so that the travel route of the vehicle 1 intersects the parking route, and when the vehicle 1 overlaps the parking route, controls the steering angle so that the travel route of the vehicle 1 follows the parking route. That is, the travel control unit 160 estimates the position and attitude of the vehicle 1, and performs feedback control on the steering angle so that the travel route of the vehicle 1 follows the parking route recorded on the map.

[0076] When performing automatic parking, the travel control unit 160 may reproduce the vehicle speed during learning travel. Further, when performing automatic parking, the travel control unit 160 may limit the vehicle speed to a predetermined value or less. This is because if the vehicle speed is high, the wheels of the vehicle 1 may slip and deviate from the route. For example, the travel control unit 160 may maintain the vehicle speed during automatic parking at 5 km / h.

[0077] Since the aforementioned position estimation unit 150 performs self-position estimation during automatic parking, it is preferable to operate the self-position estimation of the position estimation unit 150 to be complementary to the self-vehicle position estimation of the travel control unit 160.

[0078] For example, in a turning section, the vehicle may deviate from the route due to slip or centrifugal force, but in the self-vehicle position estimation of the travel control unit 160, it is difficult to detect a deviation that does not appear in the steering angle or the rotation speed of the wheels.

[0079] FIG. 9 is a diagram for explaining an example of a process of specifying a deviation when the vehicle 1 deviates from the parking route. As shown in FIG. 9, for example, when the vehicle 1 is traveling on the arc AB and there are feature points at points A and B, the difference in the azimuths of points A and B, that is, the magnitude of the angle APB, is constant if the vehicle 1 is at point P on the arc APB according to the theorem of inscribed angles.

[0080] However, when the vehicle 1 is at a point Q on the arc AQB outside the arc APB, the difference in the azimuths of point A and point B, that is, the magnitude of the angle AQB, is smaller than the angle APB. Also, the angular difference between the angle AQB and the angle APB corresponds to the magnitude of the deviation. Thus, by capturing the feature points before and after the vehicle 1 with the camera 2 and evaluating their azimuths, when the vehicle 1 deviates from the parking route, the direction and magnitude of the deviation can be determined.

[0081] Therefore, for example, when the travel control unit 160 estimates that slip has occurred from the rotation speed of the wheels of the vehicle 1, or in a turning section with a small turning radius, it evaluates that the reliability of its own vehicle position estimation is low, and may use the estimated value of the position estimation unit 150. In this case, the travel control unit 160 may acquire the data (position and orientation) of the self-position estimation of the position estimation unit 150, overwrite its own vehicle position estimation data with the data of the self-position estimation, and ensure the reliability of subsequent vehicle control.

[0082] That is, the travel control unit 160 may backup its own vehicle position estimation with the self-position estimation of the position estimation unit 150. Also, during automatic parking, the travel control unit 160 may not use its own vehicle position estimation data, acquire the self-position estimation data of the position estimation unit 150, and perform feedback control of the vehicle 1 using only the self-position estimation data.

[0083] Returning to FIG. 4, the description of the parking support ECU 100 will be continued. The map generation unit 120 stores, in the storage unit 170 as part of the map data, the feature points whose coordinates are specified by the space recognition unit 140 during learning travel. This process of storing in the storage unit 170 as part of the map data is called registering on the map. The map generation unit 120 is one of the map generation means. Since the map is generated by the storage unit 170 storing various data, it can be said that, in addition to the map generation unit 120, the storage unit 170 is also included in the map generation means.

[0084] In addition, the map generation unit 120 registers the parking route on the map. The parking route may be, for example, a parking route based on the estimated vehicle position calculated by the driving control unit 160 based on the observed values of the rotation speed and steering angle of the wheels of the vehicle 1. That is, the map generation unit 120 may register, as it is, the parking route generated by the driving control unit 160 and graphically represented in the form of a polyline from an overhead view on the map.

[0085] The map generation unit 120 of the present embodiment receives, from the driving control unit 160 at the end of the learning drive, a parking route graphically represented in the form of a polyline from an overhead view, and reconstructs it into a plurality of sections connecting between the parking start position and the parking position. Each section is either a straight section that goes straight with the steering angle zero or a turning section that turns with a constant steering angle, and approximates the polyline-shaped parking route. This may be referred to as the generation of a parking route approximating the learning drive route. For the sections having the parking start position or the parking position as endpoints, approximation is performed with the constraint condition that the parking start position and the parking position are not changed.

[0086] Then, the map generation unit 120 registers the driving distance of each section and the steering angle of the turning section on the map. In this way, by simplifying the polyline-shaped route into a small number of sections and registering it on the map, the steering angle control and vehicle speed control during automatic parking are simplified, and the data amount of the map is also reduced. In addition, the swaying of the vehicle body during automatic parking is reduced, so the satisfaction of the passengers is improved.

[0087] That is, the map generation unit 120 of the parking support ECU 100 according to the present embodiment does not completely reproduce the parking route during the learning drive for the automatic parking route, and the parking position reproduces the parking position during the learning drive. This is because generally, the passenger does not evaluate the position of the vehicle 1 on the parking route so much, but evaluates the accuracy of the parking position.

[0088] In addition, since objects around the parking route may change their positions, even if the parking route during learning driving is reproduced exactly, it is not guaranteed that the vehicle will not hit an obstacle. Therefore, in this embodiment, the parking support ECU 100 ensures that the vehicle 1 does not collide with an obstacle or the like by means of an automatic braking system. That is to say, reproducing the parking route during learning driving does not produce any particular effect, and it can be said that the necessity is lower compared with reproducing the parking position.

[0089] In this embodiment, it is described that the map generation unit 120 registers the feature points around the parking route and generates the parking route. However, regarding the generation of the parking route, another functional unit may perform it. For example, the parking support ECU 100 may include a route generation unit as a functional unit separately from the map generation unit 120.

[0090] In this case, the map generation unit 120 may register the feature points around the parking route and register the parking route generated by the route generation unit. Or, the route generation unit may generate the parking route and register the generated parking route in the map. In other words, the map generation unit 120 and the route generation unit may cooperate to generate a map.

[0091] (Process of map generation) Next, the process of map generation will be described. As a name, the manual driving that travels from the parking start position to the parking position by manual driving and stores the parking start position and the parking position is called complete learning driving. Complete learning driving is the same as what has been conventionally called learning driving. On the other hand, in the learning-type automatic parking method according to this embodiment, there are multiple types of learning driving in a form that does not pass through the parking start position. Therefore, in order to distinguish them, the learning driving that travels from the parking start position to the parking position is called complete learning driving for distinction. Hereinafter, the process of map generation will be described by taking the case of complete learning driving as an example.

[0092] In this embodiment, the process of generating a map by complete learning driving can be divided into the following processes (sub-processes).

[0093] P001: Start Point Processing (Registration of GPS Coordinates at the Starting Position) P002: Collection Processing (Collection of Feature Point Information, Route Information such as Steering Angle and Movement Amount, Detection Images, etc.) P003: End Point Processing (Receiving Completion of Learning Driving and Notifying Analysis) P004: Analysis Processing (Arrangement of Parking Routes) P005: Proposal Processing (Detection and Presentation of Parking Position Candidates) P006: Registration Processing (Registration of Parking Position and Route)

[0094] The above process of map generation is managed by the state management unit 110 as a whole, but the occupant is also involved in the process. Each functional unit shown in FIG. 4 performs corresponding processing under the management of the state management unit 110. In addition, the occupant may perform learning driving multiple times during the process of generating a map. For this reason, each functional unit may execute the corresponding processing multiple times. Note that each functional unit does not necessarily execute all processes corresponding to learning driving. For example, in the learning driving where a parking position is added without starting from the parking start position, the map generation unit 120 does not execute the first start point processing. Details of each process will be described below.

[0095] (P001: Start Point Processing) The start point processing is the processing performed at the start point (parking start position) of the learning driving. The map generation unit 120 acquires the GPS coordinates at the start position from the navigation device 40 at the start position (start point) of the learning driving so as to identify the map to be used from the GPS coordinates during automatic parking, and registers it in the map. In addition, the map generation unit 120 performs necessary initialization processing so that feature points can be extracted, coordinates can be identified, and route information can be collected.

[0096] (P002: Collection Processing) The collection processing is the processing performed on the route between the start position and the parking position. The map generation unit 120 extracts feature points and identifies coordinates, and collects information on the feature points (including coordinates). In addition, the map generation unit 120 collects route information such as steering angle, vehicle speed, gear position, moving direction, and moving distance. In addition, the map generation unit 120 stores detection images and detection data on the route and at the parking position.

[0097] (P003: Endpoint Processing) The endpoint processing is the processing performed at the end point (parking position) of the learning drive. The state management unit 110 starts the endpoint processing when the gear position becomes P. The state management unit 110 notifies the occupant to wait while analyzing the data. After the above notification by the state management unit 110, the map generation unit 120 starts the next analysis process. Note that the state management unit 110 may cause the map generation unit 120 to start the analysis process while the notification message is being output. Thereby, the state management unit 110 can shorten the waiting time felt by the occupant.

[0098] Among the processes of this map generation, the processes of P001 to P003 are the processes during the learning drive. Also, the processes from P004 onward are the processes after the learning drive. During the learning drive, the map generation unit 120 may prioritize data collection while avoiding processes with a large load, and perform processes with a large load after the learning drive.

[0099] For example, in the collection process P002, the map generation unit 120 only stores the detection images used for spatial recognition, and the spatial recognition itself may be performed by the spatial recognition unit 140 during the analysis process. In the present embodiment, the map generation unit 120 only collects the information of the feature points in the collection process, and performs the selection of the feature points and the registration of the information of the feature points after the learning drive. Or, the map generation unit 120 may register the information of the feature points during the collection process, and delete the information of the unnecessary feature points during the analysis process after the learning drive to reduce the data volume.

[0100] (P004: Analysis Processing) The analysis process is mainly the process performed by the map generation unit 120. During the endpoint processing, the map generation unit 120 receives the parking route graphically represented as a polyline by the travel control unit 160. During the analysis process, the map generation unit 120 approximates the received parking route with a straight line section that goes straight with a steering angle of zero and a turning section that turns with a constant steering angle.

[0101] Then, a large amount of data describing a large number of broken lines is replaced by a small amount of data describing a small number of sections. As a result, the route information is aggregated into the travel distance of each section and the steering angle of the turning section. The start point and end point of each section are called endpoints. When the coordinates of the endpoints of each section are determined, the travel distance of each section is determined. Therefore, instead of omitting the travel distance of each section from the data, the coordinates of the endpoints of each section may be included in the data.

[0102] The above process is a route calculation process for setting the parking route from the start position of the learning run to the parking position to generally follow the route of the learning run. Therefore, existing route calculation methods can be applied to the above process. For example, the map generation unit 120 may set a plurality of endpoints on the route of the learning run and calculate a parking route composed of straight sections and turning sections that reaches the parking position via the plurality of endpoints from the parking start position.

[0103] Note that the process of simplifying the parking route is not essential. That is, the map generation unit 120 may register the parking route graphed as a broken line by the travel control unit 160 as the parking route as it is. In that case, the route calculation process for setting a route generally following the route of the learning run is unnecessary.

[0104] The map generation unit 120 may perform the analysis process only once in one learning run, but may also perform the analysis process multiple times. For example, when an additional parking position different from the parking position of the learning run is added in the following proposal process, the map generation unit 120 repeats the analysis process and calculates a parking route to park at the additional parking position. That is, when a parking position different from the parking position of the learning run is set after the learning run, the map generation means generates another route from the parking start position of the learning run to the other parking position and registers the generated other route in the map. Then, a map having a plurality of routes from one parking start position to a plurality of parking positions can be generated, and the parking position can be selected according to the generated map.

[0105] (P005: Proposal Process) The proposal process is a process in which the state management unit 110 proposes to the passengers of the vehicle 1 to additionally register a parking position. For example, the map generation means (memory unit 170) may collect detection information around the vehicle during learning driving, and based on the detection information, set another parking position different from the parking position of the learning driving. Then, generate another route from the automatic parking route to the other parking position, and register the generated other route in the map. Then, since a plurality of parking positions and parking routes can be registered in one learning drive, the labor of the passengers can be reduced. During the proposal process, the state management unit 110 may propose additional registration when the space recognition unit 140 detects a parking available space. Also, the state management unit 110 may propose additional registration regardless of the detection by the space recognition unit 140. Hereinafter, setting another parking position by detection is called automatic setting, and setting another parking position regardless of detection is called manual setting. Here, the former automatic setting will be described, and the latter manual setting will be described later.

[0106] The space recognition unit 140 may detect a parking available space at the parking position. Also, the space recognition unit 140 may detect a parking available space in the middle of the parking route. That is, the detection for automatic setting may be executed during the collection process, during the proposal process, or across both periods. For example, if detection images and detection data are saved during the above-described collection process, it is also possible to detect a parking available space in the middle of the parking route during the proposal process. Also, the processing of the saved data may be included in the above-described analysis processing.

[0107] FIG. 10 is a diagram for explaining an example of space recognition in the proposal process. For example, when the vehicle 1 is going straight from the parking start position 19 to the turning-back position 21, the sonar system detects the distance to the side obstacles (wall surfaces) with the side sonars 3a and 3b. The space recognition unit 140 analyzes the result of the detection, and when there is no obstacle up to a distance 1.4 times the vehicle width of the vehicle 1 and the space is continuous for a length 1.2 times the vehicle length of the vehicle 1, it may be estimated that there is a parking available space on the side of the vehicle 1.

[0108] In addition, as shown near the parking position 22 in FIG. 10, after the vehicle 1 parks at the parking position 22, the space recognition unit 140 may estimate that there is a parking space between the parking position and the left wall surface in the same manner as described above. Further, the space recognition unit 140 may detect the motion parallax from the images stored for detection without relying on the side sonar, identify the visible range of the road surface by analyzing the distribution of the parallax, and estimate that there is a parking space.

[0109] When a second parking position (a parking position different from the parked position) is automatically set by detection, as will be described later, an additional parking position may be proposed, approved by the occupant, and registered on the map, or may be automatically registered without approval. Alternatively, it may be negatively notified to obtain tacit approval. For example, when it is determined in learning driving that there is a parking space for two vehicles in the garage, the second parking position may be automatically set without notifying the occupant. Alternatively, the fact that the second parking position has been registered together with the first parking position is notified by a message on the screen, and the registration is maintained unless the occupant cancels the registration. By doing so, when automatically parking, even if the first parking position is blocked, the route can be changed to the second parking position to continue automatic parking. The fact that the second parking position has been automatically set may be notified when the vehicle cannot park at the first parking position, or may be notified after automatic parking.

[0110] FIG. 11 is a diagram showing an example of an image for proposing an additional parking position in the proposal process. For example, when there is a space for the host vehicle to fit on the left side of the parked vehicle 1, the state management unit 110 may display an image for proposing an additional parking position as shown in FIG. 11. Specifically, the state management unit 110 controls the notification unit 180 to superimpose and display a semi-transparent image showing a virtual image parked adjacent to the overhead image output to the HMI device 20, and display a message such as "Do you want to register the left space as an additional parking position?" to seek approval or inquire of the occupant. Alternatively, the same message may be output by voice to seek approval.

[0111] The image of vehicle 1 on the right side of FIG. 11 is the model of the host vehicle. The other parts are the bird's-eye view images generated by projective transformation of the camera images. For example, the notification unit 180 semi-transparently superimposes the model of the host vehicle on the left side of the host vehicle in the bird's-eye view image so that a scene where two vehicles are parked in a garage can be imagined.

[0112] The proposed additional parking position may be set based on the detection information. Specifically, the state management unit 110 may set the above additional parking position by reflecting the distance between the vehicle 1 parked during learning driving and the surrounding objects (for example, a wall). For example, as shown in FIG. 11, the state management unit 110 sets the additional parking position parallel to the parked vehicle 1, and makes the distance therebetween the same as the distance between the parked vehicle 1 and the right wall 203. Further, the distance from the additional parking position to the back wall is made the same as the distance from the vehicle 1 to the back wall. In this way, the state management unit 110 can uniquely determine the additional parking position by reflecting the distance between the vehicle 1 parked during learning driving and the surrounding objects at the additional parking position. Since the parking position of the vehicle 1 reflects the distance that the occupant feels safe and the preference of the occupant, by reflecting the distance between the parking position of the vehicle 1 and the surrounding objects at the additional parking position, the additional parking position becomes preferable for the occupant.

[0113] Also, for example, the state management unit 110 may display an image indicating the automatically set additional parking position on the touch panel of the navigation device 40 so that the occupant can move the image of the additional parking position by operating the touch panel. In this way, when the automatically set additional parking position is not preferable, the occupant can manually adjust the additional parking position to suit their preference.

[0114] In the above case, for example, after the state management unit 110 displays an image proposing an additional parking position and then directly accepts the pressing of the approval button by the occupant, the map generation unit 120 registers the proposed additional parking position. Also, when the state management unit 110 accepts an operation by the occupant to move the additional parking position and then accepts the pressing of the approval button, the map generation unit 120 registers the additional parking position changed by the occupant's operation. In the former case, although the additional parking position to be registered is an automatically set position, since the occupant manually approves it, it can also be said to be a manual setting.

[0115] To distinguish the method of setting the additional parking position only through such a proposal process from the method of setting the additional parking position during additional learning driving, it is called the manual setting of the additional parking position. Also, the proposal process may be renamed as the proposal means. The proposal means proposes to the occupant to add a parking position when the vehicle parks during learning driving, and based on the operation or approval of the occupant, sets another parking position. Then, the map generation means generates another route leading to the other parking position and registers the generated other route on the map. In this way, according to the manual setting, it is possible to register the additional parking position without actually parking, and the setting can be manually changed to suit the preference, so it is simple and suitable.

[0116] Also, as described above, during learning driving, detection information around the vehicle may be collected, a parking position candidate may be automatically set based on the detection information, and the parking position candidate may be proposed to the vehicle occupant. Then, at the time of being proposed to the occupant, since it is generally set to a good parking position, the occupant may slightly modify and approve it, or approve it as it is. In this way, by using the automatic setting in combination with the manual setting, it is possible to assist the occupant in setting or save the trouble of manual adjustment.

[0117] So far, the manual setting for manually adjusting and setting the automatically set additional parking positions has been described. However, it is also possible to perform manual setting without automatic setting. That is, on the premise that the proposing means allows the occupant to set the additional parking positions, the detection process for detecting the parkable space may be omitted. For example, the state management unit 110 may simply display a query such as "Do you want to register another parking position?" on the HMI device 20 or the like and query the occupant. If the answer to the query is Yes, the state management unit 110 may set the additional parking position according to the operation of the occupant received by the HMI device 20.

[0118] When the detection process is not performed, a plurality of parking position candidates based on the layout of a general parking lot may be presented and the occupant may be allowed to make selections and operations. For example, if the occupant's answer to the query is Yes, the HMI device 20 may display an overhead image with parking frames indicating the additional parking position candidates arranged around the front, rear, left, and right of the image of the vehicle 1. Then, the occupant selects the direction for setting the parking position on this overhead image. Also, the state management unit 110 may prevent the parking frames in the direction where nearby obstacles are detected from being displayed. For example, in the arrangement as shown in FIG. 11, parking frames are not displayed on the right side and the rear of the vehicle 1. When the occupant touches one of the displayed parking frames, the other parking frames disappear, and the touched parking frame moves following the movement of the finger. The mechanism that allows the occupant to operate the position of the frame and confirm it with the approval button may be the same as the above-described method.

[0119] Here, when the addition of the parking position is approved in the proposing process, the execution order of the process for generating the map is as follows: P001: Starting point process → P002: Collection process → P003: End point process → P004: Analysis process → P005: Proposing process → P004: Analysis process → P005: Proposing process → P006: Registration process. This is because the second analysis process is executed to calculate the parking route to the additional parking position.

[0120] The following describes the second analysis process. FIGS. 12 and 13 are diagrams for explaining an example of the second analysis process. The route RA in FIG. 12 (hereinafter also referred to as the parking route RA) is the route traveled during learning driving. The learning driving starts at the parking start position 23 and ends at the parking position 27 (hereinafter also referred to as the first parking position 27), and the first analysis process and proposal process are performed after parking. For example, in the second analysis process, the map generation unit 120 branches from the parking route RA of the learning driving and generates a route RB (hereinafter also referred to as the parking route RB) for parking at an additional parking position 32 (hereinafter also referred to as the second parking position 32). If the route referred to when generating the route is the reference destination route, it can be said that the map generation means generates another route (for example, the parking route RB) based on the reference destination route (for example, the parking route RA). The reference destination route may be a set route or a route registered on the map. By using such an existing route, the time required for route calculation can be shortened.

[0121] Generally, the second parking position 32 is often set parallel to the first parking position 27. For this reason, when generating the parking route RB, the map generation unit 120 may use the parking route RA and use a parallel translation of a part of the parking route RA as the parking route RB.

[0122] As shown in FIG. 12, the parking route RA is composed of a straight section 23-24, a turning section 24-25, a turning section 25-26, and a straight section 26-27.

[0123] Since the route data is an aggregate of the data of the sections constituting the route, generating another route based on the reference destination route is nothing but generating the data of the sections of another route based on the data of the sections constituting the reference destination route. Specifically, when generating another route based on the reference destination route, the map generation means generates the data of the sections of another route by copying, processing, or sharing the data of the reference destination sections. By doing so, it is possible to shorten the time for generating the route or reduce the amount of route data.

[0124] In the second analysis process, the path for parking at the second parking position 32 is defined as the parking path RB. The map generation unit 120 may define the section of the parking path RA as the section of the parking path RB by translating it by the distance between the first parking position 27 and the second parking position 32. For example, as shown in FIG. 13, the map generation unit 120 translates the sections 24-25, 25-26, and 26-27 of the parking path RA to obtain the sections 28-29, 29-31, and 31-32 of the parking path RB. Further, the map generation unit 120 defines the straight-line section between the end point 24 and the end point 28 as 24-28.

[0125] The data of the translated section inherits the data of the original section. That is, the map generation unit 120 sets the data of the sections 28-29, 29-31, and 31-32 to be the same as the data of the sections 24-25, 25-26, and 26-27, respectively. In other words, the data of the reference section is copied to be the data of the section to be generated. Also, the map generation unit 120 sets the distance between the first parking position 27 and the second parking position 32 as the data of the distance of the straight-line section 24-28. Since it is a straight-line section, the data of the steering angle is set to zero. As a result, the map generation unit 120 completes the setting of the parking path RB that starts at the end point 23, passes through the end points 24, 28, 29, and 31, and parks at the second parking position 32. Since the parking path RB branches from the parking path RA at the end point 24, the end point 24 is called the branch point. Also, the path that starts at the end point 24, passes through the end points 28, 29, and 31, and ends at the end point 32 is called the branch path.

[0126] A branch road is a type of route. When a branch road is connected to a route registered on a map starting from the parking start position or a set route, it forms another route from the parking start position to another parking position. That is, by registering the branch road on the map, another route from the parking start position to another parking position may be registered. For example, in the registration process described later, the map generation unit 120 may register a branch road (parking route RB) that starts from the end point 24 (branch point), passes through 28, 29, 31, and parks at the second parking position 32, on the map. Then, starting from the end point 23 (parking start position), by branching at the end point 24 (branch point) from the parking route RA that parks at the first parking position 27, it is possible to park at the second parking position 32. Or, the parking route RB may be set as a route starting from the end point 23 (parking start position) and added to the map as another independent route from the parking route RA. In the latter case, the data of the first straight section 23-24 of the parking route RB is generated by copying (copying) the data of the first straight section 23-24 of the parking route RA.

[0127] When adding a branch road or a parking route to the map, the map generation unit 120 adds a link to the parking route. Here, a link refers to an address or offset that specifies the destination when jumping the data reading position. The link added to the parking route jumps to either the data of another route stored in the same map as the map storing the route or the data of another route stored in a map different from the map storing the route. That is, when a link is added to the parking route, when the state management unit 110 sequentially reads the data of the parking route from the storage unit 170 and performs automatic parking, by jumping the reading address according to the link indicating the start address of the data of another route or the branch road, the parking route and the parking position can be changed.

[0128] In the former case, a link that jumps to the data of section 24-28 of the parking route RB is added to the data of section 23-24 of the parking route RA. In the latter case, the map generation unit 120 adds a link to the parking route RA, and also adds a link that jumps to the data of section 24-25 of the parking route RA to the data of section 23-24 of the parking route RB generated by copying the data of section 23-24 of the parking route RA. This is an example of the case where data of a reference section is copied and then processed. When links to the other route are added to both the routes RA and RB, when the state management unit 110 reads the data of the parking route from the storage unit 170, it can jump to the data of the other parking route regardless of which parking route data it starts reading from.

[0129] Note that the link may be specified by an absolute address or a relative address.

[0130] For example, when the data of the parking route RA is registered in the map data first and the data of the branch road is additionally registered, the address of the data of section 23-24 of the parking route RA has been determined in advance. When the data of the branch road is additionally registered later, at that time, the starting address of the data of section 24-28 of the branch road is determined. Therefore, the map generation unit 120 may calculate the difference in addresses (address offset) and set it as a link to the data of section 23-24 of the parking route RA.

[0131] When adding a branch road to the map, the state management unit 110 must start reading the route from the parking route RA when reading the route from the storage unit 170. Then, the state management unit 110 may select whether to jump to the data of the branch road or continue reading the data of the parking route RA when reading the link.

[0132] Also, when registering the parking route RB separately from the parking route RA instead of registering the branch route, the map generation unit 120 may register the parking route RB in a separate, independent map from the map in which the parking route RA is registered. By doing so, when one map corresponds to one parking route (parking position), it becomes easier to manage the map. Since all the map data is on the ROM, it is also possible to set a link to jump to another map, and even if jumping to another map, there is no delay in processing.

[0133] Next, the data structure of the route data will be described. FIG. 14 is a diagram showing an example of the structure of the route data. In FIG. 14, the data structures of the routes RA and RB are illustrated. As shown in the figure, the route data is an aggregate of section data.

[0134] The section data consists of 3 words. A word is a unit of data having one address. One word may be, for example, 16 bits or 32 bits. The route data is included in the map data. The map data is written in an electrically rewritable ROM (for example, the ROM 102 of the parking support ECU 100). The map data can be accessed in word units.

[0135] The first word Ax of the section data indicates the steering angle and the traveling direction. The second word Dx of the section data indicates the traveling distance of the section. The third word of the section data is a link. If the link is 0, it indicates that it is the last section. If it is 1 (+1), it indicates that there is no branch and the next section continues. If it is other than 0 and 1, it indicates the relative address (offset) of the branch destination.

[0136] For example, the third word L1 of section 23-24 of route RA is the offset to the first word of section 24-28 of route RB. Also, the third word L2 of section 23-24 of route RB is the offset to the first word of section 24-25 of route RA. Usually, the first word of the next section is at the next address, so the offset in the case of no branch is +1. Therefore, the link of 1 may be called the implicit link in the case of no branch.

[0137] When the map generation unit 120 reads a link Lx that is not 0 or 1 from the storage unit 170, if it branches, it adds Lx to the address, and if it does not branch, it increments the address by 1. That is, the links in the section where there is a branch omit the implicit link (+1) when there is no branch. Also, since link = 0 indicates the final section, when the distance specified by the second word Dx has been traveled, the vehicle may stop and the automatic parking may be terminated.

[0138] In the example of FIG. 14, the map generation unit 120 generates the data of the route RB using the data of the route RA. First, the map generation unit 120 copies the leading section of the route RA and sets it as the leading section 23 - 24 of the route RB. Next, the map generation unit 120 inserts a straight section 24 - 28 after the section 23 - 24. The second word D5 of the straight section 24 - 28 is set based on the distance between the first parking position 27 and the second parking position 32 in FIG. 13. Then, when the map generation unit 120 copies the data of the sections 24 - 25, 25 - 26, and 26 - 27 of the route RA and pastes them as the data of the sections 28 - 29, 29 - 31, and 31 - 32 of the route RB respectively, the data of the route RB is completed.

[0139] In this way, by using the data of the route RA, the map generation unit 120 can generate the route RB with a small amount of calculation. In the example of FIG. 14, since the data of the route RA and the data of the route RB are independent, if the target parking position is the end point 27, the data of the route RA may be read, and if the target parking position is the end point 32, the data of the route RB may be read. Whichever route is started to be read, the route can be changed midway.

[0140] Note that the data structure of the route is not limited to the above. FIG. 15 is a diagram showing another example of the data structure of the route. For example, the map generation unit 120 may generate the data of a section of another route (route RB) by sharing a part of the data of a section of the reference destination (route RA). By sharing the data of the section, the data amount of the route can be reduced.

[0141] The example in FIG. 15 corresponds to a configuration in which a branch is added to the data of path RA. Therefore, regardless of the target parking position, the state management unit 110 starts reading from the data of the head section 23-24 of path RA. When the vehicle 1 reaches the end of the section, the state management unit 110 determines whether to branch or not. If it branches and causes the vehicle 1 to travel on path RB, it jumps to the data of section 24-28 according to the link of section 23-24 (offset = 10). The vehicle 1 travels straight by the distance indicated by D5 under the control of the travel control unit 160.

[0142] Then, since the vehicle 1 reaches the position of endpoint 28 in FIG. 13, afterwards, it may travel on a path obtained by translating sections 24-25 to 26-27 of path RA in parallel. Therefore, the state management unit 110 jumps the read address to the data of section 24-25 according to the link of section 24-28 (offset = -11). As shown in FIG. 14, the data of sections 28-29 to 31-32 of path RB is the same as the data of sections 24-25 to 26-27 of path RA. Therefore, when the address is jumped to the data of sections 24-25 to 26-27 of path RA and the same data is read, the vehicle 1 travels on a congruent path at a position translated by the distance indicated by D5.

[0143] As described above, when the second parking position can be set parallel to the first parking position, the map generation unit 120 can reduce the calculation amount of path generation and the amount of data to be handled by using the data of the existing section. Also, when the second parking position is not parallel to the first parking position and there is an angular difference, a turning section may be inserted as the first section of the branch, and the angular difference may be compensated in the turning section so that the data of the sections of the first path can be used.

[0144] In the above, an example in which the map generation unit 120 stores the data of each section of the path at consecutive addresses has been shown. However, the map generation unit 120 may arrange the data of each section in a dispersed manner. In this case, the data of the section requires a link to the next section when not branching. Therefore, the data of the section may include a data field indicating the number of links.

[0145] A data field indicates a range assigned to one piece of data. The data field may be some bits within one word. Also, the data field may span multiple words. Also, the number of links may be increased so that branching can occur in three or more directions from one branch point. Also, multiple branch points may be provided to configure a parking path that allows parking at three or more parking positions.

[0146] Next, the second proposal process will be described. After the completion of the second analysis process, the state management unit 110 performs the second proposal process together with the notification unit 180. When there is a parking available space at a location not set as a parking position in the second proposal process, the state management unit 110 may display a message such as "Do you want to register another parking position?" on the HMI device 20 or the like and query the passenger.

[0147] Also, the state management unit 110 may simply display a message such as "Do you want to register another parking position?" on the HMI device 20 or the like and query the passenger without causing the space recognition unit 140 to perform space recognition. If there is no other space where another parking position can be set, or if it is not possible to add a parking position due to restrictions such as the data capacity of the map, the state management unit 110 may display a message such as "The registration of the parking position is terminated" on the HMI device 20 or the like and notify the passenger without querying.

[0148] If the passenger's answer to the query is Yes, the same process as when the passenger's answer was Yes in the first proposal process may be repeated, so the description is omitted. If the passenger's answer is No, the state management unit 110 displays a message such as "The addition of the parking position is terminated" on the HMI device 20 or the like to notify the passenger and proceeds to the registration process.

[0149] (P006: Registration Process) In the registration process, the map generation unit 120 registers the data of the feature points in the map together with the data of the parking route. Hereinafter, registering the data of the feature points in the map is referred to as registering the feature points. Since the map generation unit 120 registers the feature points in the map to detect the movement of the vehicle 1 on the parking route, it is advisable to focus on registering the feature points that are advantageous for detecting the movement. For example, when the vehicle 1 moves or turns on the route, the feature points that move significantly on the camera image are more sensitive than the feature points with a small movement amount and are advantageous for detecting the movement of the vehicle 1.

[0150] Since the map generation means collects information on the feature points around the vehicle during the learning drive, when registering another route in the map, it is advisable to register in the map at least the information on the feature points located near another parking position. For example, when the parking position during the learning drive and the parking position during the automatic parking are different, the feature points near the parking position change their positions more significantly on the camera image than the feature points far from the parking position. Therefore, registering the feature points near the parking position preferentially can align the parking position more accurately.

[0151] Focusing on registering the feature points near a specific position can also be described as increasing the score. For example, the map generation unit 120 may increase the score in the area corresponding to the normal direction of the end point of the straight section or the area corresponding to the tangent direction of the end point of the turning section to focus on registering the feature points. Since the area where the score should be set increases when adding a parking position, the map generation unit 120 may set different scores depending on whether a parking position is added or not.

[0152] Figures 16 to 18 are diagrams showing an example of the registration process of the feature points. In Figure 16, the areas Ax (x = 1 to 6) indicated by the ellipses are the areas where the score is increased in the map in which the route RA in Figure 12 is registered. Ax in Figure 18 is the area where the score is increased in the map in which only the route RB in Figure 13 is registered. Ax in Figure 17 is the area where the score is increased in the map in which the routes RA and RB in Figure 13 are registered.

[0153] The necessity of additional registration of feature points is explained. For example, consider a map in which the feature points in the path RA in FIG. 12 and the area Ax in FIG. 16 are registered, and a path RB is additionally registered without additional registration of feature points. In that case, for example, when switching at the end point 29, since there are no feature points in the traveling direction, the position in the left-right direction and the posture (orientation) of the vehicle body may become inaccurate. Therefore, it is advisable to arrange (allocate) many feature points also in the area A7 which is the area existing in the tangential direction at the end point 29.

[0154] The feature points behind the parking position 32 are also important for accurately controlling the position in the left-right direction and the posture of the vehicle body. Therefore, the map generation unit 120 may arrange many feature points also in the area A8 existing in the traveling direction of the section 31 - 32. In this way, if the map generation unit 120 additionally arranges feature points in response to the addition of the parking path, the distribution of the areas where many feature points are arranged will be like Ax in FIG. 17, and no matter which of the parking positions 27 and 32 is selected, the vehicle 1 can perform automatic parking with high accuracy.

[0155] In addition, when the total number of feature points registered in one map is determined, if the number of parking positions increases, the density of the feature points will decrease, so there is a risk that the accuracy of parking will deteriorate. Therefore, the map generation unit 120 may generate different maps for each parking position and make the feature points registered for each map different. For example, register the feature points shown in Ax of FIG. 16 in the map in which the path RA is registered, and register the feature points shown in Ax of FIG. 18 in another map in which the path RB is registered. By doing so, the decrease in the density of the feature points can be suppressed, so the accuracy of parking can be maintained. In other words, during automatic parking, the position estimation unit 150 can maintain the accuracy of self-position estimation.

[0156] Note that even if the map is divided into two, the data volume does not necessarily double. For example, consider the case where a map M1 including the feature point Ax and the path RA in FIG. 16 and a map M2 including the feature point Ax and the path RB in FIG. 18 are generated individually. In this case, the map generation unit 120 mutually sets a link that jumps from the end point 24 of the map M1 to the end point 24 of the map M2 and a link that jumps from the end point 24 of the map M2 to the end point 24 of the map M1. Thereby, the state management unit 110 can select a parking path at the end point 24 (branch point).

[0157] Since a part of the data of the path RA and the data of the path RB are the same as shown in FIG. 14, they may be compressed as shown in FIG. 15 to reduce the data volume. However, since the data volume of the path is not originally large, there is not much difference even if it is compressed. On the other hand, many of the feature points registered in the map M2 overlap with the feature points registered in the map M1. Since the number of feature points is large and the data volume is large, if there are many overlapping feature points, the utilization efficiency of the memory capacity deteriorates. Therefore, the map generation unit 120 may store the feature points existing in the overlapping regions A1, A2, A4, and A5 in the two maps in shared data to avoid duplication.

[0158] When the state management unit 110 reads the map M1, it may read the feature point data dedicated to the map M1 (feature points existing in the regions A3 and A6) and the shared data (feature points existing in the regions A1, A2, A4, and A5). Further, when the state management unit 110 reads the map M2, it may read the feature point data dedicated to the map M2 (feature points existing in the regions A3 and A6) and the shared data (feature points existing in the regions A1, A2, A4, and A5).

[0159] Note that as described above, when there is shared data and dedicated data, the shared data and the dedicated data may be read out and merged on the RAM, or they may be accessed without distinction while the shared data and the dedicated data are placed in different regions on the ROM.

[0160] Also, when the registration process is completed, the state management unit 110 may or may not notify the passenger that the registration process has ended. This is because there is no need to request approval or operation from the passenger when the registration process ends.

[0161] For example, when an IG-OFF operation is performed during the execution of the registration process by the map generation unit 120, the state management unit 110 may request a power supply circuit (not shown) to maintain the power supply to the parking support ECU. In that case, when the registration process by the map generation unit 120 ends, the state management unit 110 may turn off the power supply without any notification. This is because the HMI device 20 and the like are not functioning due to the IG-OFF operation, so notification cannot be made. Also, if the state management unit 110 is in the IG-ON state when the registration process ends, it may display on the HMI device 20 or the like "Map registration has been completed" or "Automatic parking can be performed on the learned route" and notify the passenger. Also, if notification could not be made even though the registration process has ended, the same notification may be made when the vehicle next turns IG-ON.

[0162] So far, in the proposal process, an example has been described in which another candidate parking position is displayed on the HMI device 20 or the like, and the parking position is set by the operation and approval of the passenger. According to such a method of manually setting an additional parking position, it is convenient because the parking position can be added without actually parking. However, it can be expected that actually parking will match the passenger's preference and set an optimal parking position with respect to surrounding objects, and if actually parked, the passenger will also be convinced. Therefore, the state management unit 110 may set the parking position by having the passenger perform a manual parking.

[0163] (Continuous learning driving) Hereinafter, continuous learning driving will be described. Continuous learning driving is learning driving that starts after parking during learning driving. That is, continuous learning driving starts from the parking position where vehicle 1 parked during learning driving. The map generation means sets the parking position when the occupant of vehicle 1 manually drives vehicle 1 out of the garage and parks at a position different from the parked position at the time of leaving the garage as another parking position. Then, another route from the parking start position of the learning driving to another parking position is generated, and the generated another route is registered on the map. For example, when there are spaces for multiple vehicles to park in the garage, multiple parking positions will be registered. However, according to continuous learning driving, there is no need to return to the parking start position to start learning driving, so multiple parking positions can be registered in a short time.

[0164] FIG. 19 and FIG. 20 are diagrams for explaining an example of continuous learning driving. For example, as shown in FIG. 19, after vehicle 1 parks at the first parking position 27 via 24, 25, 26 from the end point 23, the state management unit 110 causes the HMI device 20 or the like to display "Do you want to register another parking position?" or the like and queries the occupant.

[0165] If the occupant's answer is Yes, the state management unit 110 causes the HMI device 20 or the like to display "Please move the vehicle and park at the next parking position." or the like and requests the occupant to perform manual parking. In response to this, as shown in FIG. 20, the occupant parks at the parking position 32 via the turning-back point 33 and the turning end point 34 from the parking position 27. Then, the state management unit 110 identifies the second parking position 32 based on the manual parking by the occupant.

[0166] In the above case, the process of generating a map proceeds from P001: Starting point processing → P002: Collection processing → P003: End point processing → P004: Analysis processing → P005: Proposal processing before the start of continuous learning driving, and a request for continuous learning driving is made during the proposal processing. After starting the continuous learning driving, the process of generating a map proceeds as P002: Collection processing → P003: End point processing → P004: Analysis processing → P005: Proposal processing → P006: Registration processing. Since the content of each process is generally the same as when manually setting without actually parking as described above, the main differences will be mainly explained.

[0167] In continuous learning driving, first, collection processing is performed. This is to collect feature point information even on the route to the second parking position 32 and register more advantageous feature points for detection in the map.

[0168] The second collection processing can be the same as the collection processing on the route to the first parking position 27. The map generation unit 120 treats the feature point information collected in the second collection processing and the feature point information collected in the first collection processing without distinction. The state management unit 110 performs the second end point processing when the gear position becomes P, and the notification unit 180 notifies the occupant to wait. The map generation unit 120 starts the second analysis processing simultaneously with the notification by the notification unit 180.

[0169] Before the second analysis processing, the map generation unit 120 receives from the driving control unit 160 the movement route (from the first parking position 27, via the end points 33 and 34, to the second parking position 32 route) between the first parking position 27 and the second parking position 32. This route is a movement route graphically represented by a broken line. Here, the map generation unit 120 identifies the relative positional relationship between the first parking position 27 and the second parking position 32 based on the movement route. However, the map generation unit 120 does not perform a process of approximating the received movement route (graphically represented by a broken line) with a straight line section and a turning section. This is because the route from the first parking position 27 to the second parking position 32 is an unnecessary route that is not used when automatically parking.

[0170] In the second analysis process, based on the relative positional relationship between the first parking position 27 and the second parking position 32 and the parking route RA for parking at the first parking position 27, the map generation unit 120 calculates a parking route RB as shown in FIG. 20. Previously, in the proposal process, the method of showing an overhead view to the passenger and allowing the additional parking position to be set on the screen and calculating the parking route from the parking start position to the additional parking position in the analysis process was explained. Among these processes, the process of calculating the parking route from the parking start position to the additional parking position can be implemented regardless of the method of setting the additional parking position. That is, when the second parking position 32 is set by continuous learning driving, the subsequent route calculation is as good as when manually set. Therefore, to avoid duplication, the explanation is omitted.

[0171] The next proposal process may be the same as the first proposal process performed before starting continuous learning driving, so this is also omitted from the explanation. Also, the last registration process is generally the same as when manually set.

[0172] When continuous learning driving is performed, in the registration process, the map generation unit 120 registers the feature point information collected in the first and second collection processes in the map. This is different from the case of manual setting where the collection process is only performed once. That is, after the learning driving, the map generation means collects information on the feature points around the vehicle when the vehicle occupant drives manually, and when registering another route in the map, at least the information on the feature points located near another parking position is registered in the map.

[0173] Based on the principle of triangulation, when identifying the coordinates of feature points, the accuracy of the coordinates is better when identifying them closer to the feature points than when identifying them at a distance from the feature points. For example, the coordinates of the feature points near the second parking position 32 are more accurate when identified during the continuous learning run for parking at the second parking position 32 than when identified during the first learning run. Also, feature points that were not detected in the first collection process may be detected in the second collection process, and the newly detected feature points may also contribute favorably to the position estimation of the vehicle on a new parking route. Therefore, it can be expected that the parking position accuracy when automatically parking at the second parking position 32 is better when registering the feature points collected during the continuous learning run than when registering the feature points collected during the first learning run.

[0174] When adding a parking position, the point that the area where more feature points should be arranged increases is the same as in the case of manual setting. Regarding the registration process of registering the newly collected feature points on the map, it can be the same as in the case of manual setting, so the explanation is omitted.

[0175] So far, an example of starting the continuous learning run immediately after the first learning run has been described, but the continuous learning run does not necessarily have to start immediately. For example, the occupant may start the continuous learning run on another day.

[0176] Specifically, after the first learning run, if the vehicle 1 has not moved and the data collected in the first collection process has not been erased, the occupant can start the continuous learning run even if there is an OFF / ON of the vehicle's IG in the middle. For example, when the vehicle 1 parks during the first learning run, the analysis process starts, but if the vehicle's IG is turned off without waiting for the end of the analysis process, the state management unit 110 performs the proposal process together with the notification unit 180 when the vehicle 1's IG is turned on next. Here, the occupant just needs to judge whether to perform the continuous learning run.

[0177] Further, the map generation unit 120 may execute up to the registration process of registering the map during the IG-OFF period, assuming that the registration of the map data by the first learning drive is approved. When performing continuous learning driving after registering the map, even if the data on the RAM collected in the first time has disappeared, based on the routes and feature points registered in the map, the map generation unit 120 can generate a map with the feature points collected in the continuous learning drive added or the route data added, and register it in the form of overwriting the previously registered map. Alternatively, a map registering the feature points collected in the continuous learning drive and the branch road starting from the end point 24 may be generated separately from the previously registered map and additionally registered. Further, the map to be additionally registered may cite the feature point group and route of the previously registered map.

[0178] Note that when performing the second route calculation, if the map by the first learning drive has been registered, the reference route referred to in the route calculation is the route registered in the map. Conversely, if the map has not been registered, the reference route is the set route not registered in the map. Therefore, the map generation means may be configured to generate another route based on the route registered in the map or the set route when generating another route to another parking position, and be able to select the reference route. In this way, by using an existing route, the time required for route calculation can be shortened.

[0179] (Branch learning drive) Next, the branch learning drive will be described. The branch learning drive is a learning drive that starts during automatic parking. The map generation means sets the parking position when the driver of the vehicle 1 switches to manual driving during automatic parking and parks at a position different from the target parking position of automatic parking as another parking position. Then, it generates another route from the start position of automatic parking to the other parking position, and registers the generated other route in the map. According to this branch learning drive, when it is impossible to park at the target parking position of automatic parking, it is possible to start the learning drive on the spot, and there is no need to return to the parking start position to start the learning drive, so the parking position can be added in a short time.

[0180] Here, an example of switching to manual driving midway and performing branch learning driving when driving along the parking route shown in FIG. 12 in automatic parking will be described. This example assumes that a complete learning drive has been performed in the past, starting learning driving at endpoint 23, passing through endpoints 24, 25, and 26, parking at parking position 27, and registering a map including parking route RA. Then, when the occupant starts automatic parking at endpoint 23 and is driving automatically according to the map, if the occupant steps on the brake to stop the vehicle and switches to manual driving to park, this becomes branch learning driving.

[0181] For example, when the vehicle is stopped during automatic driving in section 23 - 24 of parking route RA and manual driving is started, the map generation unit 120 may use the position where the vehicle stopped on section 23 - 24 as the branch point. For example, if the occupant stops vehicle 1 at endpoint 24, endpoint 24 becomes the branch point. Alternatively, the map generation unit 120 may compare the manual parking route with the automatic parking route, identify the position where the manual parking route diverges from the automatic parking route, and use that as the branch point. For example, after stopping on section 23 - 24 to abort automatic parking and starting manual driving, if vehicle 1 goes straight beyond endpoint 24 as in route RB in FIG. 13, the map generation unit 120 may use endpoint 24, which is the point where the routes diverge, as the branch point. Hereinafter, it will be described assuming that the branch point is endpoint 24. For example, the map generation unit 120 additionally registers the branch road starting from the branch point 24 in the map.

[0182] In the continuous learning driving described above, the vehicle departs from the first parking position and parks at the second parking position, and the vehicle 1 does not travel along the route that goes straight from the parking start position to the second parking position. Also, in manual setting, the vehicle does not park at the second parking position in the first place. Therefore, in continuous learning driving and manual setting, the route to park at the second parking position is generated based on the route to park at the first parking position. On the other hand, in branched learning driving, since the vehicle 1 travels along the route that goes straight from the parking start position to the second parking position, there is no need to generate an additional route based on the existing route. That is, the map generation unit 120 may register the route traveled manually from the branch point 24 as it is as a branch route on the map. Alternatively, a route calculation process may be performed to approximate the route of the learning driving and set a route that generally follows it, and a branch route composed of a small number of sections may be registered on the map.

[0183] However, even in the case of branched learning driving, similar to the cases of continuous learning driving and manual setting, a route to park at the second parking position may be generated based on the route to park at the first parking position and registered on the map. That is, depending on the method of setting the second parking position, there is no need to change the method of route calculation, and the route to park at the second parking position may always be generated based on the route to park at the first parking position. In the case of branched learning driving, the route to park at the first parking position is a route registered on the map, but in the case of continuous learning driving, the route to park at the first parking position has been set but is not registered on the map. Therefore, when generating another route, the map generation means generates another route based on the reference route and enables the reference route to be selected from the route registered on the map or the set route. In this way, by using the existing route, the time required for route calculation can be shortened.

[0184] In addition, when the driver of the vehicle 1 switches from automatic parking to manual driving, the vehicle's driver is prevented from being notified, or the notification is not given, that a different route will be registered on the map than when starting manual driving from the parking position. The reason is that when switching from automatic parking to manual driving, the purpose is often to avoid oncoming vehicles or obstacles, and it is rare for the purpose to be parking at a different location. For this reason, it is better for the state management unit 110 to suppress the notification or not give the notification so as not to trouble the driver as much as possible.

[0185] For example, when the driver stops the vehicle 1 and then cancels automatic parking and starts manual driving, it should not be necessary to inquire about adding a parking route or setting a parking position. If an inquiry is made, unnecessary notifications will be given when the purpose is avoidance, and the notifications will trouble the driver. Therefore, the state management unit 110 causes the notification unit 180 to output an image displaying "A manual parking route can be registered", etc., and starts the collection process so that the driver can add a parking position when desired. The notification when starting manual operation should be limited to displaying a message in text on the screen of the HMI device 20 and not giving an audio notification. That is, when the parking support device interrupts automatic parking and starts manual driving, it silently starts the learning driving process.

[0186] In the above case, it can be said that the state management unit 110 starts the learning driving process speculatively because it causes the map generation unit 120 to start the learning driving process without asking the driver whether to add a parking position.

[0187] Also, when the map generation unit 120 detects that the gear position has become P in the end point process, it may start the processes after the analysis process without notifying the driver and execute them in the background without asking for the driver's approval until the registration process. That is, all processes including map registration may be executed speculatively. In such a case, the state management unit 110 may notify after the completion of the registration process or when the IG is turned on next that a parking position has been added and that the last registered parking position can be deleted.

[0188] As described above, when switching from automatic parking to manual parking, it is more often for other purposes such as avoiding obstacles, unloading luggage, or getting people off, rather than for the purpose of adding a parking position. Therefore, the state management unit 110 may notify in a way that allows it to be left alone, rather than in a way that requires approval or instruction to proceed.

[0189] For example, the state management unit 110 displays text such as "Do you want to register the position where manual parking was last performed?" on the HMI device 20 or the like. And if Yes is not input within 5 seconds, the map generation unit 120 may delete the map of the branch road. In that case, when the passenger does not want to add the parking position, the passenger does not need to do anything.

[0190] Also, the map generation unit 120 may register and retain it with "unapproved" appended to the map. In this case, the state management unit 110 may display a list of unapproved maps when manually parking next time and request approval. Also, the map generation unit 120 may perform operations such as automatically deleting the map with the oldest registration date among the unapproved maps when the number of registered maps increases.

[0191] (Additional learning driving) Next, the additional learning run will be described. The additional learning run is a learning run that starts after automatic parking. The map generation means sets, as another parking position, the parking position when the vehicle 1 is parked by automatic parking, the occupant of the vehicle 1 drives out manually and travels, and parks at a position different from the parking position of the automatic parking. Then, another route from the parking start position of the automatic parking to another parking position is generated, and the generated another route is registered on the map. When there are spaces in the garage that can accommodate multiple vehicles, one parking position may be registered, and after trying automatic parking, the next parking position may be registered. In this case, according to the additional learning run, there is no need to return to the parking start position to start the learning run, so the next parking position can be registered in a short time. The additional learning run may be regarded as starting the continuous learning run at the parking position of the automatic parking instead of the parking position of the learning run, or the branching learning run may be regarded as starting at the parking position of the automatic parking instead of during the automatic parking.

[0192] The notification of the additional learning run may be the same as the notification of the continuous learning run. For example, in the example of FIG. 12, when the vehicle 1 automatically parks at the parking position 27, the state management unit 110 displays a message for querying, such as "Do you want to register another parking position?", on the HMI device 20 or the like. In this case, if the occupant's answer is Yes, the state management unit 110 displays a message such as "Please move the vehicle and park at the next parking position." on the HMI device 20 or the like and requests the occupant to perform manual parking.

[0193] In response, it is assumed that the occupant drives out manually and parks at the parking position 32 as in the example of FIG. 19. In this case, the map generation unit 120 identifies the second parking position 32 based on the manual parking by the occupant. In the case of the additional learning run, the method of setting the branch point and the process of generating the route of the branch road may be generally the same as those of the continuous learning run, but the data used as the basis for route generation is different. The data used as the basis for route generation in the case of the additional learning run is the data of the automatic parking, similar to the case of the branching learning run.

[0194] In route generation after the additional learning drive, based on the positional relationship between the first parking position and the second parking position and the route for parking at the first parking position, a route for parking at the second parking position is generated. In the case of continuous learning drive, the route data in the memory is the route data before registration. In the case of additional learning drive, the route data in the memory is the route registered on the map. However, since the content of the data is the same, there is no need to distinguish between the two.

[0195] Note that if there is an error in vehicle control during automatic parking and there is a difference between the target parking position (the end point of the route) registered on the map and the actual parking position, this may be compensated for in route generation. For example, the map generation unit 120 first calculates the positional relationship between the first parking position and the second parking position from the route of the additional learning drive. By adding the data of the difference between the actual parking position and the target parking position to the data of this positional relationship, the accumulation of errors can be suppressed. Therefore, the storage unit 170 may save the data of the difference between the actual parking position and the target parking position when automatically parking in preparation for additional learning drive. If this is done, the map generation unit 120 can also suppress the accumulation of errors even when performing additional learning drive on a day different from the day of automatic parking. The explanation of other processing for route calculation based on the second parking position is omitted because it overlaps with the explanation in the case of continuous learning drive.

[0196] The processing of the feature points in the additional learning drive may be the same as the processing of the feature points in the continuous learning drive. As an application when adding a parking position by learning drive, the map generation unit 120 may exclude the feature points registered in the existing map from the target of the collection process. Alternatively, the map generation unit 120 may exclude the feature points registered in the existing map from the processing target among the feature points collected in the collection process during the registration process. By reducing the processing target in this way, the processing time can be shortened.

[0197] Previously, the applications of dividing the map by the parking route RA and the parking route RB and the application of combining the maps into one were described. In the case of additional learning driving, it is advisable to additionally register a map of the branch road separately from the map of the parking route RA. In the case of additional learning driving, the branch road is often not registered, so it is easier to delete the map by separating the map of the branch road.

[0198] For example, as shown in FIG. 16, there is a map that has a route RA for parking at the parking position 27 and registered feature points in regions A1 to A6. When adding a map that has a route RB for parking at the parking position 32 and registered feature points in regions A1, A2, A4, A5, A7, and A8 as shown in FIG. 18, the two maps can be handled independently. In order to handle branches, it is necessary to change the links of the existing maps, but since the feature points and the like do not need to be changed, the amount of map change can be minimized.

[0199] If the maps are not separated, the feature points registered in the map will spread over the range of regions A1 to A8 in FIG. 17, and the number of feature points will be larger than when registering a single parking route. In the example of FIG. 17, parking is in the same garage, but in the case of additional learning driving, it is often the case that temporary parking is made at a completely different position outside the garage and no additional parking positions are registered, so the additionally registered feature points are often wasted.

[0200] When the maps are not separated and the data volume of the map is kept constant, in order for the map generation unit 120 to register the feature points collected in the second learning drive in the map, it is necessary to select the feature points already registered in the map and delete some of them. Then, there are demerits such as an increase in the processing amount and a lengthening of the processing time, or a deterioration in the parking accuracy due to deleting the feature points registered in the map. Also, when no additional parking positions are registered, the added feature points become unnecessary, but the deleted feature points cannot be restored, so the deterioration in accuracy continues. When the map generation unit 120 generates a separate map of the branch road, there are no demerits such as an increase in the processing time or a deterioration in the accuracy in exchange for an increase in the data volume.

[0201] So far, the method for generating a map including a plurality of parking positions has been described. When performing automatic parking using the map, a parking position is selected by some means. For example, when starting automatic parking or during automatic parking, the passenger may be allowed to select the parking position. However, it is better to assign priorities to the parking positions in advance. In this way, the parking support ECU 100 can automatically select a parking position based on the priorities, so the passenger does not need to give an instruction during automatic parking. The means for determining the priorities can be arbitrary. For example, the order in which the parking positions are registered can be used as the priority, or the passenger can manually determine the priorities. In the following example of automatic parking, the priorities are assigned in the order from the closest to the parking start position.

[0202] (Automatic parking process) Hereinafter, an example of the process executed by the parking support ECU 100 during automatic parking will be described. FIG. 21 is a flowchart showing an example of the process executed by the parking support ECU 100 according to the embodiment during automatic parking. FIG. 22 is a diagram for explaining an example of automatic parking executed by the parking support ECU 100 according to the embodiment.

[0203] Hereinafter, as an example of the process executed by the parking support ECU 100 according to the embodiment, an example of automatic parking using a map including a plurality of parking positions will be described. The map generated during learning driving stores a route as shown in FIG. 22.

[0204] The map generated during learning driving stores a route R-1 for parking at the parking position PP-1, a route R-2 that branches off from the route R-1 at the branch point B-1 and parks at the parking position PP-2, and a route R-3 that branches off from the route R-2 at the branch point B-2 and parks at the parking position PP-3.

[0205] Assuming the number of stored routes is M, in the example of Fig. 22, M = 3. In the automatic parking example of Fig. 22, when traveling on route R-N (N = 1 to 3), it travels with the parking position PP-N (N = 1 to 3) as the target parking position. And when reaching the branch point B-N (N = 1 to 2), if it cannot park at the parking position PP-N (N = 1 to 2), it changes the route by setting N = N + 1. However, when N = M (M = 3), if it cannot park at the parking position PP-M, it automatically stops and ends the automatic parking.

[0206] The flowchart of Fig. 21 is managed by the state management unit 110. Also, the parameters (M, N) used in the process are managed by the state management unit 110. Also, the progress of each step and the jump from step to step are determined by the state management unit 110.

[0207] First, the state management unit 110 reads a map from the storage unit 170 (step S001). The storage unit 170 can store a plurality of maps, and the map to be read is selected by the state management unit 110. For example, the state management unit 110 obtains the position information of the own vehicle from the navigation device 40 via the in-vehicle LAN, compares it with the start position coordinates of each map, and may select the map with the closest start position.

[0208] Also, the state management unit 110 sets the initial values of the parameters (M = 3, N = 1) based on the information recorded in the map read from the storage unit 170 (the number of parking positions = 3).

[0209] The space recognition unit 140 constantly detects whether there are obstacles in the traveling direction of the vehicle 1 (step S002). For example, the space recognition unit 140 detects the front based on the image of the front camera 2c. The state management unit 110 determines whether there are no obstacles within a predetermined range in the traveling direction of the vehicle 1 based on the detection result. Note that the means for obstacle detection may be a sonar or a short-range radar (not shown).

[0210] If an obstacle is detected (step S002: Yes), the state management unit 110 instructs the travel control unit 160 to automatically stop the vehicle 1 (step S003). That is, if an obstacle is detected at the start of automatic parking, the vehicle 1 does not start, and if it is detected while the vehicle is running, the vehicle automatically stops. If no obstacle is detected (step S002: No), the process proceeds to step S004.

[0211] The state management unit 110 instructs the travel control unit 160 to make the vehicle 1 travel along the route R-N (step S004). Even if the vehicle automatically stopped in step S003, if no obstacle is detected (step S002: No), the vehicle 1 resumes travel.

[0212] Next, the state management unit 110 determines whether it has reached the branch point B-N (step S005). If it has not reached the branch point B-N (step S005: No), the process returns to step S002 and the loop of steps S002 - S005 is repeated.

[0213] If it has reached the branch point B-N (step S005: Yes), the state management unit 110 determines whether the parking position PP-N is available for parking (step S006). For example, when the vehicle is at the branch point B-1, the state management unit 110 may determine that parking is possible if the space recognition unit 140 detects the left front of the vehicle 1 based on the image of the left side camera 2a and all the detection points of the parking position PP-1 among the feature points registered in the map are detected.

[0214] Also, if none of the plurality of feature points located behind the parking position PP-N among the feature points registered in the map are detected, the state management unit 110 may determine that the parking position PP-N is not available for parking.

[0215] In addition, the state management unit 110 may detect the direction of the parking position PP-N using a side sonar (not shown), and if the distance to the nearest detected obstacle corresponds to the distance to the wall behind the parking position PP-N, the state management unit 110 may determine that parking is possible at the parking position PP-N. Alternatively, the state management unit 110 may temporarily stop the vehicle 1 at the branch point B-N and inquire of the passenger whether parking is possible at the parking position PP-N and have the passenger input an answer.

[0216] If parking is not possible (step S006: No), the state management unit 110 determines whether the route being traveled is the last route registered on the map (step S007). In the example of FIG. 21, since the number M of routes is 3, if it is not the last route (N = 3), the process proceeds to step S009. If the vehicle is traveling on the last route R-3, the process proceeds to step S008.

[0217] Step S008 is the process when parking is not possible at any of the parking positions registered on the map. The state management unit 110 instructs the travel control unit 160 to stop the vehicle 1 (step S008), and ends this process.

[0218] For example, a notification process may be added to step S008. In this case, the state management unit 110 may display on the HMI device 20 or the like a message such as "Since parking is not possible at any of the registered parking positions, the automatic parking is terminated." to notify the passenger. Further, the state management unit 110 may add a step of confirming that the gear position has become P to ensure that the vehicle 1 does not move after the automatic parking is terminated.

[0219] If it is not the last route in step S007 (step S007: No), the state management unit 110 changes the route to be traveled from route R-N to route R-N + 1 by setting N = N + 1 (step S009). Step S009 is the process when the route is changed and the automatic parking is continued.

[0220] For example, if the route during driving is Route R-1 (N = 1), the state management unit 110 branches from Route R-1 to Route R-2. If the route during driving is Route R-2 (N = 2), it branches from Route R-2 to Route R-3. After branching, it returns to step S002. Steps S002 to S009 are a loop of processes corresponding to the case of repeating branching, and in this map, there is a possibility of branching up to 2 times at most.

[0221] If it was possible to park in step S006 (step S006: Yes), the state management unit 110 determines whether there is an obstacle in a predetermined range in the traveling direction of the vehicle (step S010).

[0222] If there is an obstacle in the traveling direction (S010: Yes), the state management unit 110 instructs the travel control unit 160 to automatically stop the vehicle 1 (step S011). Also, if the obstacle in the traveling direction disappears (step S010: No), the state management unit 110 instructs the travel control unit 160 to resume traveling on Route R-N (step S012).

[0223] In addition, the state management unit 110 determines whether the vehicle 1 has reached the parking position PP-N (step S013). If it has not reached the parking position PP-N (step S013: No), the state management unit 110 continues traveling until it reaches the parking position PP-N.

[0224] If it has reached the parking position PP-N in step S013 (step S013: Yes), the state management unit 110 automatically stops the vehicle 1 (step S014) and ends the automatic parking control. Note that the state management unit 110 may further add a step of activating the parking brake before the end of control to ensure that the vehicle 1 does not move after the end of automatic parking.

[0225] The flowchart of FIG. 21 inserts a process of automatically stopping vehicle 1 when an obstacle is detected. However, the state management unit 110 may perform the same process when the occupant operates the brake as when an obstacle is detected. Further, the state management unit 110 may resume the automatic parking of vehicle 1 when the occupant releases the brake.

[0226] In FIG. 21, when it is not possible to park at any of the registered parking positions on the map, an example is described in which the state management unit 110 instructs the travel control unit 160 to stop vehicle 1 and end the automatic parking. At that time, the state management unit 110 may cause the HMI device 20 or the like to display a message such as "Since it is not possible to park at any of the registered parking positions, please park manually." via the notification unit 180 to prompt the occupant to perform manual parking. Further, the notification means (the state management unit 110 and the notification unit 180) may further notify the occupant to register the manually parked position on the map. This notification means includes the state management unit 110 and the notification unit 180, and may also include the HMI device 20. When the occupant performs manual parking in response to this notification, it becomes branch learning driving.

[0227] This notification regarding the additional registration of the parking position may be performed not only when it is not possible to park at all the parking positions. In the case of FIG. 22, it is advisable to notify before the vehicle gets stuck ahead of the branch point B-2. For example, when the occupant of the vehicle stops the vehicle or when the vehicle cannot park at the parking position registered on the map, the notification means proposes to the occupant of the vehicle to switch from automatic parking to manual driving for parking. By doing so, the occupant can start branch learning driving from any position at any time without waiting for the automatic parking to get stuck. At the time of the proposal, the occupant may be notified that the parking position can be additionally registered by manual driving. By doing so, even an occupant who is unfamiliar with the parking support function can register the parking position additionally.

[0228] For example, when there are parking spaces for multiple vehicles in a garage, before the parking support device determines that it is impossible to park at all parking spaces and automatically stops the vehicle, it is considered that the driver may reach the same determination earlier. Therefore, for example, a button for selecting to switch to manual driving and park is displayed starting from the point when it is determined that parking at the first target parking space is impossible. Also, before the space recognition unit 140 starts detecting the parking space, since the driver may determine that parking is impossible, the button may be displayed when the driver stops the vehicle. By doing so, the driver can immediately start manual driving, and can smoothly add the parking space to the registered ones.

[0229] The means for instructing the start of manual parking is not limited to a button, and may be an operation of the brake, steering wheel, turn signal, or hazard lamp. Also, when starting manual parking, it is not necessarily required to stop the vehicle. For example, when the driver operates the steering wheel during automatic parking, deviates from the parking route, and parks at another position, the travel from the position where the operation of the steering wheel started to the parking position is regarded as branch learning travel, the parking route is additionally registered, and when automatically parking next time, the additionally registered parking route may be used.

[0230] Hereinafter, focusing on the process of handling data, the control of automatic parking will be described. For example, the data processing of automatic parking may be expressed as follows. As shown in FIG. 14 described above, the data of the route stored in the map is an aggregate of section data. The state management unit 110 reads out the section data and instructs the travel control unit 160 to travel based on the values of the steering angle and travel distance recorded therein. The state management unit 110 repeats this for each section, and travels along the recorded parking route.

[0231] For example, the state management unit 110 reads out the section data using a pointer. The section data has a link indicating the start address of the data of the next section. When the state management unit 110 reads out the data of the next section by following this link, it can read out the data of a series of parking routes.

[0232] When configured in this way, the branching process corresponds to the process of changing the link to be followed. For example, assume that the data of a section has a link indicating the starting address of the data of the next section and a link indicating the starting address of the data of the first section of a branch path. When the state management unit 110 reads the section data by following the latter link, it can obtain the steering angle when traveling on the branch path and the value of the travel distance of that section.

[0233] Then, the state management unit 110 successively follows the links to the next section, and when it reaches the link "0" indicating the end point, that is the final section. When the state management unit 110 finishes running the final section, it stops the vehicle and ends the automatic parking.

[0234] Here, at a branch point, the state management unit 110 selects whether to branch depending on whether the parking position connected to the route is parkable or not. Therefore, it can be said that the state management unit 110 functions as a route selection means for selecting a route at a branch point. Also, since the determination of whether parking is possible or not is performed by the space recognition unit 140, the space recognition unit 140 may be included in the route selection means.

[0235] In summary, the embodiment of the present disclosure includes a map generation means for registering a map including a route from a parking start position to a parking position during learning driving, and a travel control means for autonomously driving a vehicle based on the map during automatic parking. The map has a plurality of routes that branch in the middle from one parking start position and lead to a plurality of parking positions. During automatic parking, the parking support device further includes a route selection means for selecting a route for the vehicle to travel from the plurality of routes. By using such a parking support device, even if the parking position planned at the start of automatic parking is blocked, due to the function of the route selection means, it is possible to change the route midway and park at another parking position.

[0236] Note that the route selection means may detect a parking position or a parking route when the vehicle approaches a branch point, and select a route according to the detection result. By doing so, the vehicle automatically selects an available parking position and parks automatically, so the passenger does not need to give an instruction for automatic parking. For example, when approaching a parking position, the space recognition unit 140 analyzes the motion parallax of the image of the parking space. If there is a larger parallax than when the floor surface of the parking space is reflected, the space recognition unit 140 determines that parking is not possible because the floor surface is not visible, and may branch to a branch road leading to another parking position.

[0237] For example, the route selection means may select a route according to the detection result of the image processing unit 130. Specifically, the image processing unit 130 detects feature points on the back of the parking space or on the floor of the parking space registered in the map. When the corresponding feature points are not detected, the image processing unit 130 determines that parking is not possible because the parking space is not visible, and may branch to a branch road leading to another parking position.

[0238] For example, the route selection means may select a route according to the detection result of an obstacle detection device such as a sonar or a radar. Specifically, the obstacle detection device detects whether there is an obstacle in the direction of the parking position. When the distance to the detected obstacle is shorter than the distance to the parking position, the obstacle detection device determines that parking is not possible because there is an obstacle, and may branch to a branch road leading to another parking position.

[0239] Alternatively, the route selection means may select a route according to the instruction of the passenger. This may be rephrased as follows: when the vehicle approaches the branch point, the route selection means makes an inquiry about the parking position to the passenger of the vehicle, and selects the route along which the vehicle travels according to the response of the passenger. That is, the user interface that allows the passenger to select the parking route may be the route selection means. By doing so, it becomes possible to change the parking position according to the preference and convenience of the passenger. Also, if the passenger is to make the selection, the detection of the parking position becomes unnecessary, so the price of the parking support device can be suppressed.

[0240] For example, the state management unit 110 may stop the vehicle 1 at a branch point and output an audio message such as "Do you want to park at the parking position on the left? Or do you want to go straight?" to the HMI device 20 or the like. In this case, the state management unit 110 may select a route by accepting the pressing of a button displayed on the touch panel by the occupant. Alternatively, a button for selecting a branch while the vehicle is running may be displayed, and if the button is pressed before reaching the branch point, the vehicle may branch, and if not, it may not branch.

[0241] In addition, the state management unit 110 may notify the occupant of the result of selecting a parking route by means of a device such as the space recognition unit 140 via the notification unit 180. Furthermore, the state management unit 110 may allow the occupant to change the route selection when the notification result and the occupant's judgment are different.

[0242] For example, consider the case where the state management unit 110 notifies that "Pass the parking position on the left and go straight." and the occupant thinks that "There is no problem with the parking position on the left and I want to park at the parking position on the left." In this case, the occupant may instruct a route change by, for example, gently turning the steering wheel to the left. Then, when the state management unit 110 receives the steering wheel operation by the occupant, it may reverse the determination, select the leftward route, and at the same time, notify that "Park on the left." At that time, the state management unit 110 may notify that the instruction has been received by vibrating the steering wheel or the like. By doing so, it is possible to prevent the occupant from steering more than necessary.

[0243] As described above, the embodiments of the present invention have been described. However, the above-described embodiments are presented as examples and are not intended to limit the scope of the present invention. This novel embodiment can be implemented in various other forms. Also, various omissions, replacements, and changes can be made without departing from the gist of the invention. In addition, this embodiment is included in the scope and gist of the invention, and is included in the invention described in the claims and the equivalent scope thereof.

Explanation of reference numerals

[0244] 1 Vehicle 2 Camera 10 Operating device 20 HMI device 30 Vehicle control device 40 Navigation device 50 Sonar ECU 100 Parking support ECU 101 CPU 102 ROM 103 RAM 104 I / O 105 IMP 106 Communication I / F 110 State management unit 120 Map generation unit 130 Image processing unit 140 Space recognition unit 150 Position estimation unit 160 Driving control unit 170 Memory unit 180 Notification unit

Claims

1. A map generation means for registering a map including a route from a parking start position to a parking position during learning driving; A driving control means for autonomously driving a vehicle based on the map during automatic parking; Comprising; The map has a plurality of routes from one parking start position to a plurality of parking positions, During automatic parking, a route selection means for selecting a route on which the vehicle travels from the plurality of routes, A parking support device further comprising; The map generation means is; After the learning driving, when another parking position different from the parking position is set, Generating another route from the parking start position of the learning driving to the another parking position, and registering the generated another route in the map, Parking support device.

2. The map generation means is; Setting, as the another parking position, a parking position when the vehicle is parked by manual driving by an occupant of the vehicle from the parking position where the vehicle was parked during the learning driving, The parking support device according to claim 1.

3. The map generation means is; Setting, as the another parking position, a parking position when the occupant of the vehicle switches the automatic parking to manual driving and drives and parks manually, The parking support device according to claim 1.

4. The map generation means is; Setting, as the another parking position, a parking position when the vehicle is parked by manual driving by an occupant of the vehicle from the parking position where the vehicle was parked during the automatic parking, The parking support device according to claim 1.

5. The map generation means is; Collecting detection information around the vehicle during the learning driving, When the vehicle is parked during the learning driving, Based on the detection information, setting the another parking position, The parking support device according to claim 1.

6. Further comprising a proposal means, The proposal means is; When the vehicle is parked during the learning driving, Proposing to the occupant of the vehicle to add a parking position, and setting the another parking position based on the operation or approval of the occupant, The parking support device according to claim 1.

7. The map generation means is; Collecting detection information around the vehicle during the learning driving, setting parking position candidates based on the detection information, The proposal means proposes the parking position candidates to the occupant of the vehicle, The parking support device according to claim 6.

8. The map generation means is; When generating the another route, Using a route registered in the map or a set route as a reference route, Based on the reference route, Generating the another route, The parking assistance device according to claim 1.

9. The data of the route is an aggregate of data of sections constituting the route, The map generation means, Based on the route of the reference destination, When generating the other route, By copying, processing, or sharing the data of the section of the reference destination, the data of the section of the other route is generated. The parking assistance device according to claim 8.

10. The data of the route has a link that jumps to the data of the other route, The link jumps to either the data of the other route stored in the same map as the map storing the route, Or, The data of the other route stored in a map different from the map storing the route. The parking assistance device according to claim 9.

11. Further comprising notification means, The notification means performs notification regarding registering the other route in the map, When the vehicle occupant switches the automatic parking to manual driving, The vehicle occupant suppresses or does not perform the notification compared to when starting manual driving from the parking position. The parking assistance device according to claim 1.

12. The map generation means, During the learning driving, collects information on feature points around the vehicle, When registering the other route in the map, Registers at least the information on feature points located near the other parking position in the map. The parking assistance device according to claim 1.

13. The map generation means, After the learning driving, when the vehicle occupant drives manually, collects information on feature points around the vehicle, When registering the other route in the map, Registers at least the information on feature points located near the other parking position in the map. The parking assistance device according to claim 1.

14. The route includes a branch point, The route selection means, When the vehicle approaches the branch point, makes an inquiry about the parking position to the vehicle occupant, Selects the route along which the vehicle travels according to the response of the occupant. The parking assistance device according to claim 1.

15. The route includes a branch point, The route selection means When the vehicle approaches the branch point, detects the parking position or the parking route, Selects the route along which the vehicle travels according to the result of the detection. The parking assistance device according to claim 1.

16. Further comprising notification means, The notification means, When the vehicle occupant stops the vehicle, or, When the vehicle cannot park at the parking position registered on the map, propose to the passengers of the vehicle to switch the automatic parking to manual driving and park. The parking support device according to claim 1.

17. When a passenger of the vehicle performs a predetermined operation, switch the automatic parking to manual driving. The predetermined operation is an operation of a predetermined button, or an operation of a brake, a steering wheel, a turn signal, or a hazard lamp. The parking support device according to claim 3.

Images