Driver assistance device, driver assistance procedure and driver assistance program

Abstract

Driving assistance device (10) which assists the driving of a vehicle (1), wherein the driving assistance device (10) comprises: a training data generator (130) that stores a route driven by the vehicle (1) according to a user's driving operation in a training driving mode and generates training data (D1) for a target route based on the route driven; and a vehicle control system (160) that causes the vehicle (1) to travel along the target route when an execution command for an automatic driving mode is received, wherein the training data generator (130) generates the target route by connecting a first point on the route or a point on the route adjacent to the first point and a second point on the route or a point on the route adjacent to the second point, in a case where a difference in the orientation of the vehicle (1) at the first point and at the second point during the training run is less than a first threshold and a distance between the vehicle (1) at the first point and the vehicle (1) at the second point during the training run is less than a second threshold.

Description

Technical area

[0001] The present disclosure relates to a driver assistance device, a driver assistance procedure and a driver assistance program. State of the art

[0002] Parking spaces around houses are generally narrower, creating a strong need for automated parking assistance. Various driver assistance systems are available for this purpose. When a driver brings a vehicle to a stop at a predetermined initial holding position, such a system performs automated driving to assist the vehicle, for example, when parking, from the initial holding position to a target parking position within a parking space. The system automatically detects the parking space and the vehicle's position, for example, using various sensors mounted on the vehicle, and automatically operates the steering, accelerator, brakes, and so on.

[0003] Patent Literature (hereinafter referred to as "PTL") 1 discloses a technology for realizing such automated driving. According to the technology, a vehicle is driven from a predetermined position outside a parking space to a target parking position within the parking space by a driving maneuver performed by a user riding on the vehicle (hereinafter simply referred to as "user"), and the driving route is stored as training data. This allows the vehicle to automatically navigate along the driving route in future parking maneuvers.

[0004] PTL 2 discloses a method for operating a driver assistance system for a motor vehicle, in which, in a learning mode of the driver assistance system, during which the motor vehicle is manually maneuvered from a starting position to a target position along at least one learning trajectory by a driver of the motor vehicle, the at least one learning trajectory is recorded and objects in an environment of the motor vehicle are detected, and in an operating mode of the driver assistance system, the motor vehicle is maneuvered at least semi-autonomously between the starting position and the target position based on the at least one recorded learning trajectory and the detected objects, wherein, in the learning mode, a drivable area for the motor vehicle is determined based on the at least one recorded learning trajectory and the detected objects.within the drivable area at least one trajectory between the starting position and the target position is determined, and the motor vehicle is maneuvered along this at least one trajectory in the operating mode.

[0005] PTL 3 discloses the provision of a parking aid that enables a vehicle to be parked more efficiently in a parking space. A parking aid device has: a learning device configured to learn a first position, which is the vehicle's position when a driver begins a parking maneuver; a second position, which is the vehicle's position when the driver changes a gear position; and a third position, which is the position when the driver ends the parking maneuver during a period of time while performing the parking maneuver; and a generating device configured to generate a route, as a target route, along which the vehicle should travel when the vehicle is automatically parked, reaching the third position from the first position via the second position based on a learning result. Citation listPatent literature PTL 1 Japanese published patent application no. JP 2017 - 138 664 A PTL 2 German published patent application no. DE 10 2017 110 020 A1 PTL 3 Japanese published patent application no. JP 2019 - 99 078 A Non-patent literature

[0006] NPL 1 Mikael Persson et al. "Lambda Twist: An Accurate Fast Robust Perspective Three Point (P3P) Solver.", ECCV 2018, pp. 334-349, published in 2018, http: / / openaccess.thecvf.com / content_ECCV_2018 / papers / Mikael_Persson_Lambda_Twist_An_ECCV_2018_paper.pdf Summary of the inventionTechnical problem

[0007] According to a driver assistance device that uses the technology disclosed in PTL 1, a single training drive by the user, which is a drive to generate the training data, enables future automatic parking maneuvers. The technology is therefore useful.

[0008] However, if the user drives the vehicle into the parking space and turns back and forth, i.e., changes the vehicle's direction of travel, the vehicle may travel unnecessary distances to change its position, i.e., to change its orientation. In other words, a driving path generated by training drives may contain unnecessary distances.

[0009] This means that in a case where the route of the training drive is used as the target route for automatic parking into the parking space, as is done by the driver assistance device disclosed in PTL 1, the vehicle has to drive unnecessary distances each time.

[0010] The present disclosure is made taking into account the above-mentioned problem and aims to provide a driver assistance device, a driver assistance procedure and a driver assistance program, each of which is capable of achieving a more suitable parking operation. Solution to the problem

[0011] The present disclosure provides a driver assistance device that supports the driving of a vehicle, wherein the driver assistance device comprises: a training data generator that stores a route driven by the vehicle according to a user's driving operation in a training driving mode and generates training data for a target route based on the route; and a vehicle control that causes the vehicle to drive along the target route when an execution command for an automatic driving mode is received, wherein the training data generator generates the target route by connecting a first point on the route or a point on the route adjacent to the first point and a second point on the route or a point on the route adjacent to the second point.in which a difference in the orientation of the vehicle at the first point and at the second point during the training run is less than a first threshold and a distance between the vehicle at the first point and the vehicle at the second point during the training run is less than a second threshold.

[0012] In another aspect, the present disclosure provides a driving assistance method that supports the driving of a vehicle, wherein the driving assistance method comprises: first processing the storage of a driving distance traveled by the vehicle according to a driving operation of a user in a training driving mode and generating training data for a target route based on the driving distance;and second processing of causing the vehicle to travel along the target route when an execution command for an automatic driving mode is received, wherein the first processing comprises generating the target route by connecting a first point on the route or a point on the route adjacent to the first point and a second point on the route or a point on the route adjacent to the second point, in a case where a difference in the orientation of the vehicle at the first point and at the second point during the training drive is less than a first threshold and a distance between the vehicle at the first point and the vehicle at the second point during the training drive is less than a second threshold.

[0013] In another aspect, the present disclosure provides a driver assistance program that supports the driving of a vehicle, wherein the driver assistance program comprises: first processing of the storage of a driving route traveled by the vehicle according to a driving operation of a user in a training driving mode and the generation of training data for a target route based on the driving route;and second processing of causing the vehicle to travel along the target route when an execution command for an automatic driving mode is received, wherein the first processing comprises generating the target route by connecting a first point on the route or a point on the route adjacent to the first point and a second point on the route or a point on the route adjacent to the second point, in a case where a difference in the orientation of the vehicle at the first point and at the second point during the training drive is less than a first threshold and a distance between the vehicle at the first point and the vehicle at the second point during the training drive is less than a second threshold. Advantageous effects of the invention

[0014] According to the driver assistance device of the present disclosure, it is possible to achieve a more suitable parking maneuver. Brief description of the drawings Fig. Figure 1 is a block diagram showing an exemplary general configuration of a vehicle; Fig. Figure 2A shows exemplary function blocks of a driver assistance device for a training driving mode; Fig. 2B shows exemplary function blocks of the driver assistance device for an automatic driving mode; Fig. 3A shows an example route (a route indicated by an arrow with a solid line) that the vehicle travels in training driving mode; Fig. Figure 3B shows an example of a target route (a route indicated by an arrow with a solid line) to be traveled by the vehicle in automatic driving mode; Fig. Figure 4 shows example training data generated by a training data generator; Fig. Figure 5 is a diagram that describes an exemplary procedure for generating the training data; Fig. Figure 6 shows an example of map data generated by the training data generator; Fig. 7 is a diagram that describes an example of processing in a position estimator; Fig. Figure 8 is a flowchart that describes an exemplary operation of the driver assistance device for generating the target route; Fig. 9A shows an example route (a route indicated by an arrow with a solid line) that the vehicle travels in training driving mode according to variant 1; Fig. 9B shows an example target route (a route indicated by an arrow with a solid line) to be traveled by the vehicle in automatic driving mode according to variant 1; Fig. Figure 10 is a flowchart describing an exemplary operation of the driver assistance device for generating the target route according to variant 1; Fig. Figure 11 shows an exemplary display of the user interface on a human-machine interface for a user selection operation for the driver assistance device according to variant 4; Fig. 12 is a diagram that schematically illustrates the processing in the training data generator according to variant 5; and Fig. Figure 13 is a flowchart that describes an exemplary operation of the driver assistance device for generating the target route according to variant 5. Description of the embodiments

[0015] A preferred embodiment of the present disclosure is described in detail below with reference to the accompanying drawings. It should be noted that components having essentially the same functions are identified by the same reference numerals in this description and the drawings, and recurring descriptions are omitted. [General vehicle configuration]

[0016] An exemplary configuration of a vehicle according to one embodiment is described below with reference to Fig. 1 described.

[0017] Fig. Figure 1 is a block diagram showing an exemplary general configuration of vehicle 1.

[0018] The vehicle 1 includes a driver assistance device 10, a vehicle drive unit 20, vehicle-side sensors 30, a vehicle-side camera 40, an external storage device 50 and a human-machine interface 60.

[0019] The vehicle drive unit 20 enables the driving, braking, and turning movements necessary for the movement of the vehicle 1 and includes, for example, a drive motor, a power transmission mechanism, a braking device, a steering device, and an electronic control unit (ECU) that controls these components. The vehicle drive unit 20 propels the vehicle 1 by, for example, generating power with the drive motor and transmitting the power to the wheels via the power transmission mechanism (e.g., a driveshaft, a differential, and a drive shaft). It should be noted that, according to the present embodiment, the operation of the vehicle drive unit 20 is controlled by the driver assistance device 10 during an automatic driving mode.

[0020] The vehicle-side sensors 30 are various sensors mounted on the vehicle 1 that detect the vehicle's driving state. These include, for example, an accelerator pedal opening sensor that detects when the accelerator pedal is opened, a steering angle sensor that detects the steering angle, an acceleration sensor that detects longitudinal acceleration acting on the vehicle 1, a torque sensor that detects torque acting on the power transmission mechanism between the wheels and the vehicle's drive motor, and a vehicle speed sensor that detects the vehicle's speed. The vehicle-side sensors 30 transmit the sensor information obtained through detection to the driver assistance device 10.

[0021] The vehicle-mounted camera 40 is an environmental sensor mounted on the vehicle 1 that monitors the vehicle's surroundings. In the present embodiment, the vehicle-mounted camera 40 is used, for example, to detect objects (typically ground-mounted objects) located around the vehicle 1 and to estimate the vehicle 1's position based on the positional relationship between the vehicle 1 and the objects surrounding it. The vehicle-mounted camera 40 consists of four cameras arranged to detect, for example, the four directions in front, behind, left, and right of the vehicle 1. The vehicle-mounted camera 40 generates camera images and outputs them to the driver assistance device 10. It should be noted that, for example, instead of the vehicle-mounted camera 40,A Light Detection and Ranging (LiDAR), radar, or ultrasonic sensor can be used as an environmental sensor to estimate the self-position of the vehicle 1.

[0022] The external storage device 50 is an additional storage device, such as a hard disk drive (HDD), a solid-state drive (SSD), or flash memory. The external storage device 50 stores, for example, training data D1 and map data D2 generated in the driver assistance device 10 when the vehicle is in a training driving mode.

[0023] The human-machine interface 60 is a user interface, such as a touchscreen, a control unit, buttons, or control keys, that receives input operations from a user traveling in the vehicle 1. The human-machine interface 60 is configured to receive input operations, such as an execution command for the training driving mode and an execution command for the automatic driving mode. The human-machine interface 60 outputs the information about the received user input operations to the driver assistance device 10.

[0024] The driver assistance device 10 is an electronic control unit that generally controls every part of the vehicle 1. The driver assistance device 10 is configured to cause the vehicle 1 to drive automatically (i.e., autonomously) by controlling the vehicle drive unit 20, while referring to the sensor information acquired by the vehicle's sensors 30 in automatic driving mode.

[0025] The driver assistance device 10 is configured to include, for example, a processor 10a (CPU), a read-only memory 10b (ROM), a random-access memory 10c (RAM), an input port (not shown), and an output port (not shown). Functions of the driver assistance device 10, which are described later, are executed by the processor 10a with reference to a control program or various data stored, for example, in the random-access memory 10c or read-only memory 10b. It should be noted that some or all of the functions of the driver assistance device 10 can be implemented by processing a digital signal processor (DSP) or a special hardware circuit (e.g., ASIC or FPGA) instead of, or in conjunction with, the processing of the CPU 10a.

[0026] The driver assistance device 10 is connected via a vehicle-internal network (e.g. a communication network that complies with the CAN communication protocol) to the vehicle drive unit 20, the vehicle-side sensors 30, the vehicle-side camera 40, the external storage device 50 and the human-machine interface 60 and can send and receive the required data and control signals to and from each other. [Configuration of the driver assistance device]

[0027] Next, an exemplary configuration of the driver assistance device 10 will be described with reference to the Fig. 2A to 7 described.

[0028] The driving assistance device 10 is configured to switch between training driving mode and automatic driving mode based on user input. Training driving mode is used to register a target route for vehicle 1 in order to drive in automatic mode. In training driving mode, the vehicle 1's movement is controlled by the user's driving input. Automatic driving mode is a mode that causes vehicle 1 to drive automatically along the target route registered during training driving mode. In automatic driving mode, vehicle 1's movement is automatically controlled by the driving assistance device 10 without the need for the user to drive.

[0029] It should be noted that the driving assistance device 10 according to the present embodiment is configured to automatically generate the target route to be driven by the vehicle 1 by cutting off an unnecessary route due to curves that are included in the driving route of the vehicle 1 in training driving mode.

[0030] In Fig. 2A and Fig. Figure 2B shows exemplary functional blocks of the driver assistance device 10. It should be noted that in Fig. 2A only shows the function blocks that work when vehicle 1 is in training driving mode, and in Fig. 2B shows only the function blocks that work when vehicle 1 is in automatic driving mode.

[0031] Fig. Figure 3A shows an example route (a route indicated by a solid line arrow) driven by vehicle 1 in training driving mode. Fig. Figure 3B shows an example target route (a route indicated by a solid line arrow) to be travelled by vehicle 1 in automatic driving mode.

[0032] Fig. Figure 3A illustrates a parking maneuver in training mode by having the user drive from position P1 outside the parking space to the target parking position P2 inside the parking space. The target path in Fig. 3B is based on the route traveled by vehicle 1 in the Fig. The training driving mode shown in 3A is generated, and most of the target route is the same as the driving route. The only changes made to the target route compared to the original driving route are the turning points.

[0033] More precisely, vehicle 1 will be at positions T1 and T2 (here from forwards to backwards) on the track in Fig. 3A turned around, but the target route in Fig. 3B is generated by cutting off unnecessary distances due to the turns of vehicle 1 on the route, and vehicle 1 is turned at positions T1' and T2'. <Während des Traingsfahrmodus>

[0034] First, a configuration of the functions of the driver assistance device 10, which operate during the training driving mode, is described.

[0035] During training driving mode, the vehicle information collector 110, the coupling computer 120 and the training data generator 130 function in the driver assistance device 10 (see Fig. 2A).

[0036] The vehicle information collector 110 acquires sensor information indicating the driving status of the vehicle 1 from the vehicle's sensors 30. The vehicle information collector 110 then transmits the acquired sensor information to the coupling computer 120.

[0037] The coupling computer 120 estimates the current position of vehicle 1 based on temporal changes in the sensor values ​​of the vehicle's sensors 30, which detect the vehicle's driving state. The coupling computer 120 calculates the amount of movement of vehicle 1 from a reference position (e.g., a starting position at the beginning of the training driving mode) based on, for example, temporal changes in vehicle speed and yaw rate displayed by the vehicle's sensors 30, and estimates the current position of vehicle 1 based on this amount of movement.

[0038] The current position of vehicle 1 estimated by the coupling computer 120 is an approximate current position of vehicle 1. The position estimate by the coupling computer 120 has low accuracy, and therefore, in the driver assistance device 10 according to the present embodiment, the estimated current position of vehicle 1 is corrected to the actual current position, which is based on camera images generated by the vehicle-side camera 40. The position estimate by the coupling computer 120 has low accuracy because, for example, 1) it is difficult to construct a complete kinematic model (e.g., a tire diameter and tread width), 2) the observed values ​​from the vehicle-side sensors 30 always contain noise, and 3) external factors (e.g., tire slippage) that cannot be observed by the vehicle-side sensors 30 strongly influence the position estimate.Additionally, due to these circumstances, the position estimation by the coupling computer 120 accumulates errors.

[0039] The training data generator 130 stores the route of the training drive and generates training data D1 from this route, which relates to the target route that vehicle 1 is to automatically travel in automatic driving mode (hereinafter referred to simply as the "target route"). For example, the training data generator 130 starts storing the route in response to a user command to start the training driving mode and stops storing the route in response to a user command to end the training driving mode. The training drive typically begins with vehicle 1 stopping at a desired position outside the parking space (P1 in the parking area). Fig. 3A) and ends with the vehicle 1 stopping at a parking destination position in the parking lot (P2 in Fig. 3A). It should be noted that the training data D1 generated by the training data generator 130 is stored in the external storage device 50.

[0040] Fig. Figure 4 presents exemplary training data D1, generated by training data generator 130. The data in Fig. The data D1a shown in Figure 4 are data that indicate the actual driving distance of vehicle 1 and a driving condition at each point of the driving distance, which are stored sequentially during the training drive, and the training data D1 are generated from the data D1a of the training drive.

[0041] Fig. Figure 5 is a diagram that describes an exemplary procedure for generating training data D1.

[0042] It should be noted that Fig. 4 and Fig. 5 the workflow of cutting off unnecessary sections before and after the turning point T1 in Fig. 3. Describe WP_1, WP_2, ..., WP_G in Fig. 4 and Fig. Point 5 represents each point along the route of vehicle 1 that was recorded during the training run. WP_1 corresponds to the starting point of the route, and WP_G corresponds to the endpoint of the route.

[0043] The training data D1 includes, for example, the target route that vehicle 1 is to travel automatically, the orientation (i.e., the direction of view) of vehicle 1 at each target point on the target route, the direction of travel (i.e., forward or backward) of vehicle 1 at each target point on the target route, and reference travel information at each target point on the target route.

[0044] The “target route” of the training data D1 is generated, for example, by deleting the unnecessary sections before and after the turning point in the data D1a of the route in the training drive and connecting the sections before and after the unnecessary sections.

[0045] More precisely, the training data generator 130 searches, according to the connection conditions, for a pair of first and second points that have the turning point of the direction of travel on the training run route between them. The connection conditions are that the difference in the orientation of vehicle 1 at the first and second points in the training run is less than the first threshold, and that the distance between vehicle 1 at the first point and vehicle 1 at the second point is less than the second threshold.When the pair of first and second points that fulfills the conditions for the connection is detected, the training data generator 130 establishes a connection between the first point, or a point adjacent to the first point on the route, and the second point, or a point adjacent to the second point on the route, and generates the target route by shortening the route at the connection point. That is, the training data generator 130 configures the connection point of the first point (or the point adjacent to the first point on the route) and the second point (or the point adjacent to the second point on the route) as a turning point for vehicle 1 on the target route, instead of the turning point on the route of the training run.

[0046] It should be noted that the first threshold for the difference in orientation (the orientation of vehicle 1 when vehicle 1 passes the points), which serves as a reference for the conditions for the connection, is, for example, a value of 0 degrees or more and 3 degrees or less. Furthermore, the second threshold for the distance, for example, a value of 0 m or more and 1 m or less, serves as a further reference for the conditions for the connection.

[0047] The training data generator 130 performs such processing because the pair of first and second points that fulfills the conditions for the connection can essentially be considered the same point that vehicle 1 passes during the training run, with essentially the same position before and after the turning point of vehicle 1 on the track. In other words, the distance between the pair of first and second points and the turning point on the track is an unnecessary distance that has no effect on the change in position of vehicle 1. It should be noted that a pair of points WP_N and WP_K on the track in Fig. 5 corresponds to the pair of the first and second points, that the conditions for the connection are met.

[0048] When the training data generator 130 generates the target route, the route between the two connected points is preferably not orthogonal to the routes before and after the two points. Therefore, for at least one of the first and second points, the training data generator 130 configures the point adjacent to the first or second point on the route as the connection target. For example, in Fig. 5. It is preferable to establish a connection between point WP_N-1 and point WP_K, rather than a connection between point WP_N and point WP_K, which are the pair that fulfills the conditions for the connection. That is, in Fig. 5. The target route is preferably configured such that vehicle 1 travels forward from point WP_N-1 to point WP_K and turns around at point WP_K (see the connection position of the training data D1 in Fig. 4).

[0049] This prevents vehicle 1 from moving laterally (i.e., in the direction of the vehicle width) at the connection point on the target track and allows vehicle 1 to travel smoothly on the target track. Furthermore, configuring point WP_N-1 as one of the connection points instead of point WP_N (i.e., moving the point consisting of the pair of first and second points on the starting point side of the track by one point closer to the starting point), as shown in Fig. Figure 5 shows that this prevents the steering angle from being excessively large when the vehicle 1 travels along the route from point WP_K.

[0050] It should be noted that the exemplary processing for generating the target route in training data generator 130 will be described later with reference to Fig. 8 is described.

[0051] The “orientation” of vehicle 1 at each target point on the target track and the “direction of travel” of vehicle 1 at each target point on the target track contained in the training data D1 are, for example, the direct mirror image of the orientation and direction of travel of vehicle 1 during the training run.

[0052] Furthermore, the “reference drive information” at each target point on the target route, contained in the training data D1, includes, for example, driving information (e.g., the steering angle and vehicle speed) that is recorded at each point on the route during the training drive of vehicle 1. This reference drive information is used as a reference (e.g., referred to as a forward coupling element) when the vehicle controller 160, which is described later, controls the vehicle drive unit 20 (e.g., the steering angle of the steering device) at each point on the route during the execution of the automatic drive.

[0053] It should be noted that the training data generator 130 preferably changes the vehicle speed at the connection position (here at point WP_K) on the target track to zero [m / sec] from the vehicle speed in the training run, together with the elements of the "reference run information" at each target point on the target track in the training data D1. This prevents a rapid change in vehicle speed when the vehicle 1 turns around at the connection position.

[0054] The training data generator 130 generates training data D1 while simultaneously generating map data D2 to estimate the current position of vehicle 1, for example, from the camera images of the vehicle's camera 40. This technique is also known as simultaneous localization and mapping (SLAM).

[0055] Fig. Figure 6 illustrates exemplary map data D2 generated by training data generator 130. Fig. Figure 6 is a bird's-eye view showing the real spatial positions of characteristic points Q in a real scene stored in the map data D2.

[0056] The map data D2 is data that stores, for each of a multitude of characteristic points in a real-world scene, a three-dimensional position of the characteristic point in real space, in conjunction with a characteristic value of the characteristic point, which was obtained from a camera image captured at the time the map data D2 was generated. The characteristic points stored as map data D2 are, for example, areas (e.g., corner areas) where characteristic image patterns are obtained from camera images of objects (e.g., a tree, a wall, or a column) that can serve as features in the real-world scene. It is important to note that the multitude of characteristic points in the map data D2 are stored in a way that allows for individual identification, e.g., by identification numbers.

[0057] The three-dimensional positions in real space of the characteristic points stored in the map data D2 are represented by a three-dimensional orthogonal coordinate system (X, Y, Z).

[0058] The characteristic magnitude of the characteristic point stored in the map data D2 includes, for example, a characteristic magnitude of the scale-invariant feature transformation (SIFT) or a characteristic magnitude of the accelerated robust features (SURF), as well as the brightness and density on the camera image. It is important to note that the characteristic magnitude data of the characteristic point stored in the map data D2 can be stored separately for each camera position and orientation when capturing the characteristic point, even for the characteristic point at the same three-dimensional position. Furthermore, the characteristic magnitude data of the characteristic point stored in the map data D2 can be stored in conjunction with the image of the object containing the characteristic point.

[0059] The training data generator 130 determines the coordinates of characteristic points in the real-world scene, for example, based on stereophotogrammetry. More precisely, the training data generator 130 reads a multitude of camera images acquired at different times and assigns the identical characteristic points that appeared together in the multitude of camera images. Next, the training data generator 130 receives information, for example, from the coupling computer 120, about the temporary position of vehicle 1 when the multiple camera images are acquired, and identifies the temporary coordinates of the characteristic points in the real-world scene using a triangulation principle. Subsequently, the training data generator 130 performs bundle adjustment, for example, by...The training data generator 130 uses the temporary position of vehicle 1 and the temporary coordinates of the characteristic points in the real-world scene as reference information and calculates the official position of vehicle 1 and the official coordinates of the characteristic points in the real-world scene to minimize reprojection errors when each characteristic point in the real-world scene is projected onto all camera images. The training data generator 130 then stores the official position of vehicle 1 as data D1a, which indicates the position of vehicle 1 during the training run. The training data generator 130 also stores the official coordinates of the characteristic points in the real-world scene as map data D2 in the external storage device 50.

[0060] It should be noted that the map data D2 can be generated in advance using LiDAR or a stereo camera instead of SLAM methods. However, the generation of map data D2 often introduces distortion depending on the measurement method. Therefore, from the perspective of avoiding a deterioration in position estimation accuracy due to distortion in map data D2, it is preferable to use the SLAM method. <Im automatischem Fahrmodus>

[0061] Next, a configuration of the functions of the driver assistance device 10, which operate during automatic driving mode, is described.

[0062] During automatic driving mode, the vehicle information collector 110, the coupling computer 120, the position estimator 140, the destination track reader 150 and the vehicle control unit 160 function in the driver assistance device 10 (see Fig. 2B).

[0063] The vehicle information collector 110 and the coupling computer 120 have the same configuration as those described above, so their description is unnecessary.

[0064] The position estimator 140 estimates the current position of vehicle 1 based on map data D2 and camera images from the vehicle's camera 40.

[0065] Fig. Figure 7 is a diagram that describes an example of processing in the Position Estimator 140. The points R1, R2, and R3 in Fig. 7 represent three characteristic points extracted from the camera image of the vehicle-mounted camera 40. Points Q1, Q2, and Q3 represent the three-dimensional positions of characteristic points R1, R2, and R3 in real space and are stored in the map data D2. RP1 represents an imaging surface of the vehicle-mounted camera 40. Point P1 represents the position of the vehicle-mounted camera 40 (i.e., the position of the vehicle 1), which is determined from the three characteristic points R1, R2, and R3 extracted from the camera image of the vehicle-mounted camera 40 and the points Q1, Q2, and Q3 stored in the map data D2.

[0066] For example, the position estimator 140 first compares the characteristic points extracted from the camera image of the vehicle-mounted camera 40 with the characteristic points stored in the map data D2, e.g., using pattern matching and feature amount search. The position estimator 140 then randomly selects several (e.g., three to six) characteristic points from among those extracted from the camera image of the vehicle-mounted camera 40 that match the characteristic points stored in the map data D2, and estimates the position of the vehicle 1 in real space based on the positions of the several characteristic points in the camera image and the three-dimensional positions of the several characteristic points in real space stored in the map data D2.At this time, the position estimator 140 calculates the position of vehicle 1 by solving a P-NP problem using a known method such as Lambda Twist (see e.g. NPL 1).

[0067] It should be noted that if the position estimator 140 compares the characteristic points extracted from the camera image of the vehicle-side camera 40 with the characteristic points stored in the map data D2, then the characteristic points stored in the map data D2 can be narrowed down to the characteristic points that match the characteristic points extracted from the camera image of the vehicle-side camera 40, based on the current position of the vehicle 1, which was estimated, for example, by the coupling computer 120.

[0068] Information about the current position of vehicle 1, estimated by the position estimator 140, is transmitted to the vehicle controller 160. It should be noted that the information about the current position of vehicle 1 estimated by the position estimator 140 includes, for example, information about the two-dimensional position (the X-coordinate and the Y-coordinate) of vehicle 1 in real space, as well as information about the orientation (i.e., the direction of view) of vehicle 1.

[0069] The target track reader 150 reads the training data D1 for the moving vehicle 1 from the external storage device 50 and transmits the training data D1 to the vehicle control unit 160.

[0070] When an execution command for automatic driving mode is received from a user, the vehicle controller 160 moves the vehicle 1 along the target route, which is indicated by the training data D1 acquired by the target route reader 150, while detecting the current position of the vehicle 1. The vehicle controller 160 then stops the vehicle 1 at the endpoint of the target route (i.e., at a position corresponding to the endpoint of the driving route in the training drive).

[0071] The vehicle control unit 160 typically controls the movement of vehicle 1 (e.g., the steering angle of vehicle 1's steering device) through feedback control based on the current position of vehicle 1, estimated by the position estimator 140, and the target points on the target route indicated by the training data D1. Such control is implemented because the initial position of vehicle 1 at the start of automatic driving mode does not always exactly correspond to the initial position of the target route specified by the training data D1. Furthermore, reproducing the route using only dead reckoning has a high probability of increasing errors, especially at low speeds and on routes with curves.Since the target route for automatic driving is generated by changes to the route of the training drive, the actual route may differ from the target route if the vehicle speed or steering angle is applied directly in the training drive.

[0072] It should be noted that the vehicle control 160 can use the reference driving information configured in the training data D1 (vehicle speed, steering angle, etc.) as a forward coupling element.

[0073] When the vehicle controller 160 instructs vehicle 1 to drive automatically, vehicle 1 does not necessarily begin its automatic journey from the position of the target track's starting point specified in the training data D1, but can begin from a position between the starting point and the endpoint of the target track specified in the training data D1. In this case, the vehicle controller 160 only needs to determine the most easily reachable target point among those between the starting point and the endpoint of the target track specified in the training data D1, based on the initial position and orientation of vehicle 1 at the start of the automatic driving mode, for example, using a predefined evaluation function. This allows vehicle 1 to drive smoothly to the target track even if its initial position differs from the target track's starting point. [Operation of the driver assistance device]

[0074] Fig. Figure 8 is a flowchart describing an exemplary process when the driver assistance device 10 (training data generator 130) generates a target route. The in Fig. The flowchart shown in Figure 8 is a process carried out by the driver assistance device 10, for example, after a training drive. It should be noted that the flowchart in Figure 8 is a process carried out by the driver assistance device 10, for example, after a training drive. Fig. 8 WP_N is used as a variable representing one of the candidate points for the pair that meets the conditions for the connection (hereinafter referred to as the "two-point pair") among the points WP_1, WP_2, ..., WP_G on a route.

[0075] In step S1, the driving assistance device 10 first sets N = 1 and configures one of the candidate points for the two-point pair as the starting point on the driving route.

[0076] In step S2, the driver assistance device 10 searches for a point that fulfills the conditions for connecting point WP_G to point WP_N+1 on the driving path, with reference to the orientation and position of vehicle 1 at point WP_N during the training run. That is, the driver assistance device 10 searches the driving path for the point that fulfills the condition for the connection and can be coupled to point WP_N in sequence from the endpoint of the driving path. This makes it possible to identify the point closest to the endpoint of the driving path first, provided that such a point exists that fulfills the condition for the connection and can be paired with point WP_N on the driving path.

[0077] In step S3, the driver assistance device 10 determines whether the point fulfilling the connection condition is detected at any point between point WP_G and point WP_N+1 on the route. If the point fulfilling the connection conditions is not detected at any point between point WP_G and point WP_N+1 on the route (NO in S3), the driver assistance device 10 continues with processing in step S4. If the point fulfilling the connection condition is detected at any point (e.g., in this case point WP_K) between point WP_G and point WP_N+1 on the route (YES in S3), the driver assistance device 10 continues with processing in step S5.

[0078] In step S4, the driver assistance device increases 10 N to N+1 for one of the candidate points for the two-point pair (i.e., it moves one of the candidate points for the two-point pair by one point towards the endpoint of the driving path) and continues with processing in step S8.

[0079] In step S5, the driver assistance device removes 10 WP_N to WP_K-1 from column WP of data D1a, which relates to the route generated in the training run, and establishes a connection between WP_N-1 and WP_K. This generates training data D1, which relates to the target route, in which the unnecessary section from the training run route is deleted (see Fig. 4 and Fig. 5).

[0080] In step S6, the driver assistance device 10 configures point WP_K as a turning point in the training data D1. More precisely, the driver assistance device 10 changes the vehicle speed for the training data D1 at point WP_K in data D1a, which refers to the route generated in the training run, to zero, since the vehicle speed at point WP_K is not zero in the training run. It should be noted that the driver assistance device 10 can also change the vehicle speed to a value close to zero at the points before and after point WP_K (e.g., WP_K+1 and WP_N-1), in order to further prevent changes in vehicle speed at the time of the turn.

[0081] In step S7, the driver assistance system increases N to K+1 for one of the candidate points for the two-point pair and continues processing in step S8. It should be noted that N is increased to K+1 because points WP_N to WP_K-1 are removed by the processing in step S5.

[0082] In step S8, the driver assistance device 10 determines whether the variable relating to one of the candidate points for the two-point pair is incremented to G-1 (N = G-1). If the variable relating to the candidate points is not incremented to G-1 (NO in S8), the driver assistance device 10 returns to step S2 and performs the same processing. If the variable relating to the candidate points is incremented to G-1 (YES in S8), the driver assistance device 10 continues with processing in step S9.

[0083] In step S9, the driver assistance device 10 saves the training data D1 generated in steps S2 to S8 in the external storage device 50.

[0084] The processing described above generates the target route, in which the unneeded section is deleted from the route of the training drive. [Effect]

[0085] As described above, the driving assistance device 10, according to the present embodiment, searches for the pair of first and second points that have the turning point in the direction of travel on the distance between them. The conditions for the first and second points to be considered are that the difference in the orientation of the vehicle 1 at the first and second points during the training drive is less than the first threshold, and that the distance between the vehicle 1 at the first point and the vehicle 1 at the second point is less than the second threshold. When the pair is detected, the driving assistance device 10 establishes a connection between the first point or the point adjacent to the first point on the distance and the second point or the point adjacent to the second point on the distance, and generates the target route for the automatic drive by shortening the distance at the connection point.

[0086] This allows vehicle 1 to drive to the target parking position without following an unnecessary route when vehicle 1 is automatically parked in the parking space during automated driving. This reduces the driving time of the automated journey and allows for optimized vehicle operation.

[0087] It should be noted that the following has also been investigated as a method for generating the target route according to conventional technology. According to this method, the route is automatically generated by storing only a start point, a turning point, and an end point of a driving path and using, for example, arithmetic processing with a scoring function to seamlessly connect the three points. However, such a method optimizes the route between the three points using a scoring score, and therefore a vehicle may pass through a location that a user would not want to pass through (e.g., grass) or an impassable location that is difficult for sensors to detect as an obstacle (e.g., a thin post, a chain, and a pond), even if the user has intentionally avoided such locations during the training drive.

[0088] However, the driving assistance device 10 according to the present embodiment basically controls the vehicle 1 in such a way that it drives along a route intended by the user, thereby avoiding the problem as with conventional technology. (Variant 1)

[0089] The above embodiment describes a case in which the training data generator 130, when generating a target path, connects a pair of two points that have different directions of travel and a turning point in the direction of travel between them. However, the training data generator 130 can also connect a pair of two points that have the same direction of travel when generating a target path, as long as the pair of two points has the turning point in the direction of travel between them and the condition for the connection is met.

[0090] Fig. 99A illustrates an exemplary route (a route indicated by a solid line arrow) driven by vehicle 1 in training driving mode according to variant 1. Fig. Figure 9B illustrates an exemplary target route (a route indicated by a solid line arrow) to be travelled by vehicle 1 in automatic driving mode according to variant 1.

[0091] Fig. 9A describes a case in which vehicle 1, while driving from position P1 outside the parking space to the target parking position P2 inside the parking space, changes its direction of travel twice due to a user's driving maneuver. When searching for a pair, the training data generator 130 determines that the pair of points WP_L and WP_M on the driving path fulfills the conditions that the difference in orientation is less than the first threshold and the distance is less than the second threshold. Therefore, in variant 1, the training data generator 130 generates the target path by connecting point WP_L-1 (the point next to point WP_L on the starting point side of the driving path) and point WP_M, as shown in Fig. Figure 9B is shown. Note that the direction of travel at point WP_L and the direction of travel at point WP_M are identical in this case (i.e., driving in reverse).

[0092] Such a configuration can also generate the target route without unnecessary sections on the training run route. In this case, however, vehicle 1 travels without turning at the connection point between the two points (the connection point can be referred to as the "non-turning point"). The user finds it inconvenient when vehicle 1 stops at the connection point between the two points during automatic travel in such a situation, as in the embodiment described above. Therefore, in variant 1, the training data generator 130 preferably sets the vehicle speed at the connection point to a non-zero value (e.g., maintains the vehicle speed from the training run) when two points with the same direction of travel are connected.

[0093] Fig. Figure 10 is a flowchart that describes an exemplary process when the driver assistance device 10 (training data generator 130) generates the target route according to variant 1. The processing of steps S1 to S9 is the same as in the flowchart in Fig. 8 described processing. The difference to the flowchart in Fig. 8 consists in the fact that this flowchart contains step Sa between step S4 and step S5.

[0094] Step Sa is executed after the pair of two points WP_N and WP_K that meet the conditions for connection has been detected and the points WP_N-1 and WP_K, which each correspond to the pair of two points, have been connected.

[0095] In step Sa, the driver assistance device 10 refers to the switching information at point WP_N and the switching information at point WP_K in the training run and determines whether the direction of travel at point WP_N and the direction of travel at point WP_K are identical or opposite to each other. If the direction of travel at point WP_N and the direction of travel at point WP_K are opposite (No in step Sa), the driver assistance device 10 configures the connection position (here point WP_K) as the turning point in the training data D1 (step S6) and continues with processing in step S7. If the direction of travel at point WP_N and the direction of travel at point WP_K are identical (Yes in step Sa), the driver assistance device 10 continues with processing in step S7 without configuring the connection position (here point WP_K) as the turning point in the training data D1.

[0096] If the connection position is configured as a turning point in the training data D1 (step S6), the driver assistance device 10 sets the target speed of the vehicle to zero at the connection position configured in the training data D1 (here point WP_K), for example, as in the embodiment described above. If the connection position is not configured as a turning point in the training data D1, the driver assistance device 10, on the other hand, sets the target speed of the vehicle to, for example, the vehicle speed during the training run at the connection position configured in the training data D1 (here point WP_K).

[0097] As described above, in some cases the target path configuration method according to variant 1 allows the configuration of a more optimal target path than the target path configuration method according to the embodiment above. (Variant 2)

[0098] In the embodiment described above, the training data generator 130 applies the same connection conditions (the first threshold for orientation and the second threshold for distance) to all candidate points when searching for the two-point pair. However, when searching for the two-point pair, the training data generator 130 can adjust the first threshold and / or the second threshold for the connection conditions described above so that they are higher the greater the distance between the candidate point and the endpoint of the travel path in the training run. In other words, the training data generator 130 can relax the connection condition for determining whether the candidate points correspond to the two-point pair for the candidate points that are farther from the target parking position.

[0099] Generally, the space around a target parking position is narrow, but there is often more space away from the target parking position. In other words, during automatic driving, if vehicle 1 is far from the target parking position, it can be guided to the target parking position without colliding with an obstacle, even if it deviates slightly from the training route. However, if vehicle 1 deviates even slightly from the training route when it is close to the target parking position, it may collide with an obstacle.

[0100] When searching for the two-point pair according to variant 2, the first threshold and / or the second threshold are set with respect to the conditions for the connection described above so that they are larger when the distance between the candidate point and the endpoint of the route in the training drive increases, thereby more actively removing unnecessary sections. (Variant 3)

[0101] In the embodiment described above, all points from the start point to the end point of the training drive's travel path are search targets for the two-point pair sought by the training data generator 130. However, when searching for the two-point pair, the training data generator 130 can exclude a point from the search targets if it has a curvature equal to or greater than the third threshold value among the points on the training drive's travel path. It should be noted that the third threshold value is, for example, a curvature that is derived from the limit of the vehicle 1's general steering angle.

[0102] When vehicle 1 travels along a specific path, the curvature of the path must generally be equal to or less than a predetermined value due to limitations on vehicle 1's steering angle. However, after changing direction, vehicle 1 is typically forced to travel along a path with a specific curvature. Therefore, connecting a point with a curvature of the predetermined value, or more than one of the two-point pair, may result in the subsequent path having excessive curvature. Consequently, vehicle 1 is unable to travel along the target path and thus deviates from it. As a result, if such a situation occurs near the target parking position, vehicle 1 could collide with an obstacle. This is because the target parking position typically has only a narrow area around it.

[0103] Limiting the search targets, as in variant 3, prevents the curvature of the target path from becoming too large.

[0104] It should be noted that the curvature determined in variant 3 at each point could be a curvature of the curve connecting the point to its neighboring point, or it could be a curvature of the curve connecting the point to the target parking position. In this case, only points around the target parking position can be excluded from the search. (Variant 4)

[0105] When the two-point pair that meets the connection condition is detected, the training data generator 130 can determine, based on a user selection operation, whether the two-point pair should be connected.

[0106] Fig. Figure 11 illustrates an exemplary display of the user interface on the human-machine interface 60 for the user's selection operation of the driver assistance device 10 according to variant 4. It should be noted that the left-hand figure in Fig. 11 represents a route of the training drive, and the right-hand illustration in Fig. 11 represents a target route suggested by the training data generator 130.

[0107] With an ad like in Fig. 11. The user can select whether to change the route to the shortcut route for each of the turning points T1 and T2, while visually confirming the user interface screen. It should be noted that Rb1, Rb2, and Rb3 are in Fig. There are 11 symbols for user input operations.

[0108] When a user drives vehicle 1 during a training run, the route is sometimes intentionally chosen by the user. For example, if the user intends to activate a monitoring sensor installed in a parking space by passing a predetermined position, configuring the target route by changing the training run route to the shortcut route would contradict the user's intention.

[0109] According to variant 4, the user can select and specify whether to configure the target path as the same path as the training run without any changes, or as a path shortened at a position of the two-point pair that fulfills the conditions for connecting points on the training run path. This allows the target path to be configured according to the user's intent. (Variant 5)

[0110] The embodiment described above describes a case in which the training data generator 130 generates a target route for automatic driving by referring only to a single training drive and optimizing that route. However, the training data generator 130 can also generate a target route by combining the data from previous training drives with the data from the last training drive.

[0111] If the parking run for the training run is only performed once, the training run route may contain unnecessary sections unless the user is experienced in driving. In this respect, it would be very advantageous to generate an optimal route by performing the training run multiple times and combining the routes. The driver assistance device 10 according to variant 5 responds to such a requirement.

[0112] Fig. Figure 12 is a diagram that schematically illustrates the processing in the training data generator according to variant 5. The left figure in Fig. Figure 12 illustrates a training run route from a specific position P1 outside the parking space to the target parking position P2 inside the parking space. The solid line shows the route of the last training run, and the dashed line shows the route of a previous training run (i.e., saved data). The figure on the right in Fig. Figure 12 shows a target route generated by combining the route of the last training run and the route of a previous training run.

[0113] For example, the training data generator 130 compares the vehicle's orientation at each point on the track of the last training run with the vehicle's orientation at each point on the track of the previous training run, searching for a point where the difference in orientation is less than the first threshold and the distance is less than the second threshold. When the point is found where the difference in orientation is less than the first threshold and the distance is less than the second threshold, the training data generator 130 considers this point as a candidate point where the track of the last training run and the track of the previous training run are combined (hereinafter also referred to as the "point of similar orientation").

[0114] Training Data Generator 130 selects either the route of the last training run or the route of a previous training run for each of the routes from the start point to the candidate point, the route from the candidate point to the next candidate point, and the route from the next candidate point to the endpoint, and generates the target route, which combines the route of the last training run and the route of the previous training runs. At this point, Training Data Generator 130 selects either the route of the last training run or the route of a previous training run by determining which of the routes is optimal for the route connecting each point, using an evaluation function based, for example, on the number of curves and the distance.

[0115] In Fig. In section 12, points Ts11 and Ts12 correspond to the candidate points where the difference in orientation is smaller than the first threshold and the distance is smaller than the second threshold. In the right-hand figure in Fig. 12 selects the travel distance of a previous training run for the route from the starting point to candidate point Ts11, the travel distance of the last training run for the route from candidate point Ts11 to candidate point Ts12 and the travel distance of a previous training run for the route from candidate point Ts12 to the endpoint.

[0116] Fig. Figure 13 is a flowchart that describes an exemplary process when the driver assistance device 10 (training data generator 130) generates the target route according to variant 5.

[0117] In step S11, the driver assistance device 10 compares the stored route A, which represents the route of a previous training drive, with the route B, which represents the route of the last training drive, and looks for the point Ts with a similar location (i.e. the point where the difference in orientation is smaller than the first threshold and the distance is smaller than the second threshold).

[0118] In step S12, the driver assistance device 10 determines whether a point of similar location Ts exists. If a point of similar location Ts exists, the driver assistance device 10 continues with processing in step S13. If no point of similar location Ts exists, the driver assistance device 10 continues with processing in step S15.

[0119] In step S13, the driver assistance device 10 selects either route A or route B as the optimal route from the starting point to the point of similar location Ts. Then, the driver assistance device 10 selects either route A or route B as the optimal route from the point of similar location Ts to the next point of similar location Ts (or to the endpoint if the next point of similar location Ts does not exist). Subsequently, the driver assistance device 10 combines the optimal routes at the point(s) of similar location Ts. The driver assistance device 10 performs this processing for all detected points of similar location Ts.

[0120] In step S14, the driver assistance device 10 presents a user with the route optimized in step S13, as well as route A and route B, so that the user can select which route he wants to accept, and continues with the processing in step S16.

[0121] In step S15, the driving assistance device 10 presents the user with both route A and route B, so that the user can select which route he wants to take over, and continues with the processing in step S16.

[0122] In step S16, the driving assistance device 10 saves the route selected by the user in the external storage device 50 as training data (i.e., the target route for automatic driving).

[0123] As described above, in some cases the target path configuration method according to variant 5 allows the configuration of a more optimal target path than the target path configuration method according to the embodiment above.

[0124] While various embodiments have been described above, it can be assumed that various changes in form and detail can be made without deviating from the spirit and scope of the invention(s) claimed now or hereafter. Commercial applicability

[0125] A more suitable parking maneuver is achieved with a driver assistance device according to the present disclosure. Reference symbol list 1 vehicle 10 Driver assistance device 20 vehicle drive unit 30 Vehicle-side sensor 40 Vehicle-side camera 50 External storage device 110 vehicle information collectors 120 coupling calculators 130 Training Data Generator 140 position estimators 150 target track readers 160 Vehicle control D1 Training data D2 map data

Claims

[1] Driving assistance device (10) which supports the driving of a vehicle (1), the driving assistance device (10) comprising: a training data generator (130) that stores a driving route traveled by the vehicle (1) according to a driving operation of a user in a training driving mode and generates training data (D1) for a target route based on the driving route; and a vehicle controller (160) that causes the vehicle (1) to travel along the target route when an execution command for an automatic travel mode is received, wherein the training data generator (130) generates the target route by connecting between a first point on the route or a point on the route adjacent to the first point and a second point on the route or a point on the route adjacent to the second point in a case where a difference in the orientation of the vehicle (1) at the first point and the second point during the training run is smaller than a first threshold value and a distance between the vehicle (1) at the first point and the vehicle (1) at the second point during the training run is smaller than a second threshold value. [2] Driving assistance device (10) according to claim 1, wherein the first point and the second point include an intermediate turning point of the direction of travel on the route of the training trip. [3] The driving assistance device (10) according to claim 2, wherein the training data generator (130) searches for a pair of the first point and the second point such that the difference in the orientation of the vehicle (1) at the first point and the second point during the training run is smaller than the first threshold value and the distance between the vehicle (1) at the first point and the vehicle (1) at the second point during the training run is smaller than the second threshold value, and, when the pair is detected, generates the target route by connecting between the first point on the route or the point on the route adjacent to the first point and the second point on the route or the point on the route adjacent to the second point, and the route at a connecting position of the first point on the route or the point on the route adjacent to the first point,and the second point on the route or the point on the route adjacent to the second point. [4] The driving assistance device (10) according to claim 1, wherein the training data generator (130) generates the target route such that the vehicle (1) changes a traveling direction at a connecting position of the first point on the traveling route or the point on the traveling route adjacent to the first point and the second point on the traveling route or the point on the traveling route adjacent to the second point. [5] The driving assistance device (10) according to claim 1, wherein the training data generator (130) sets the target vehicle speed to zero at an inflection point which is a connecting position of the first point on the travel route or the point on the travel route adjacent to the first point and the second point on the travel route or the point on the travel route adjacent to the second point, and stores the target vehicle speed in the training data (D1) in association with the target route. [6] The driving assistance device (10) according to claim 1, wherein the training data generator (130), when generating the target route for at least one of the first point of the travel route and the second point of the travel route, sets the point adjacent to the at least one of the first point and the second point as a connection target, so that a route between two points to be connected is not orthogonal to routes before or after the two points. [7] The driving assistance device (10) according to claim 6, wherein the training data generator (130) connects, when generating the target route, between a point adjacent to the first point and located on one side of a starting point on the driving route and the second point. [8] Driving assistance device (10) according to claim 1, wherein: the first threshold value is a value of 0 degrees or more and 3 degrees or less; and the second threshold is a value of 0 m or more and 1 m or less. [9] Driving assistance device (10) according to claim 3, wherein the training data generator (130) determines, when the pair is recognized, whether a direction of travel of the vehicle (1) at the first point and a direction of travel of the vehicle (1) at the second point are identical or opposite to each other, stores the connection position as a turning point in the training data (D1) in connection with the target route when the direction of travel of the vehicle (1) at the first point and the direction of travel of the vehicle (1) at the second point are opposite to each other, and stores the connection position as a non-turning point in the training data (D1) in connection with the target route when the direction of travel of the vehicle (1) at the first point and the direction of travel of the vehicle (1) at the second point are identical. [10] Driving assistance device (10) according to claim 3, wherein the training data generator (130) sets the first threshold value and / or the second threshold value in the search for the pair such that it increases with increasing distance on the route from an end point of the route to a possible point for the pair. [11] The driving assistance device (10) according to claim 3, wherein the training data generator (130), when searching for the pair, excludes a point among the points on the route that has a curvature equal to or greater than a third threshold value from possible points for the pair. [12] The driving assistance device (10) according to claim 1, wherein the vehicle controller (160) causes the vehicle (1) to travel along the target route and stops the vehicle (1) at a position corresponding to an end point of the target route when receiving the automatic driving mode execution command. [13] The driving assistance device (10) according to claim 3, wherein the training data generator (130) determines whether the target route should be configured as a route in which the travel route is shortened at the connecting position or as a route of the travel route without change based on a selection operation of a user when the pair is recognized. [14] The driving assistance device (10) according to claim 1, wherein the driving assistance device (10) outputs the driving distance traveled by the vehicle (1) according to the user's driving operation and a distance in which the first point and the second point are connected to a display device. [15] The driving assistance device (10) according to claim 1, further comprising a position estimator (140) that estimates a current position of the vehicle (1) based on sensor information of an environmental sensor that monitors an environment of the vehicle (1), wherein the vehicle controller (160), upon receiving the automatic driving mode execution command, performs feedback control of the vehicle (1) based on the current position of the vehicle (1) estimated by the position estimator (140) and target points on the target route, so that the vehicle (1) travels along the target route. [16] Driving assistance device (10) according to claim 1, wherein the driving assistance device (10) is used for parking the vehicle (1) in a parking space. [17] Driving assistance method that supports the driving of a vehicle (1), the driving assistance method comprising: first processing of storing a travel distance traveled by the vehicle (1) according to a user's driving operation in a training driving mode and generating training data (D1) for a target distance based on the travel distance; and second processing of causing the vehicle (1) to travel along the target route when an execution command for an automatic driving mode is received, wherein the first processing comprises generating the target route by connecting between a first point on the travel route or a point on the travel route adjacent to the first point and a second point on the travel route or a point on the travel route adjacent to the second point in a case where a difference in the orientation of the vehicle (1) at the first point and the second point during the training run is smaller than a first threshold and a distance between the vehicle (1) at the first point and the vehicle (1) at the second point during the training run is smaller than a second threshold. [18] Driving assistance program that supports the driving of a vehicle (1), the driving assistance program comprising: first processing of storing a travel distance traveled by the vehicle (1) according to a user's driving operation in a training driving mode and generating training data (D1) for a target distance based on the travel distance; and second processing of causing the vehicle (1) to travel along the target route when an execution command for an automatic driving mode is received, wherein the first processing comprises generating the target route by connecting between a first point on the travel route or a point on the travel route adjacent to the first point and a second point on the travel route or a point on the travel route adjacent to the second point in a case where a difference in the orientation of the vehicle (1) at the first point and the second point during the training run is smaller than a first threshold and a distance between the vehicle (1) at the first point and the vehicle (1) at the second point during the training run is smaller than a second threshold.

Images