Vehicle control device, vehicle, vehicle control method, and vehicle control program

Abstract

To allow for accurately determining whether or not a detected approaching object should be excluded from collision determination.SOLUTION: A vehicle control device provided herein is designed to be mounted on a vehicle and comprises an acquisition unit configured to acquire detection information on an obstacle detected around the vehicle, and a sensor control unit configured to perform collision determination to evaluate possibility of collision with the obstacle. The sensor control unit generates approaching object information on an obstacle approaching the vehicle and detection point cloud information on a set of detection points representing a stationary obstacle on the basis of the detection information, and excludes the approaching object from collision determination if the detection point cloud has a shielding effect that shields the vehicle from the approaching object, where the shielding effect is evaluated using a gap threshold based on a vehicle width.SELECTED DRAWING: Figure 11

Description

[Technical field]

[0001] The present disclosure relates to a vehicle control device relating to collision prevention, a vehicle equipped with a collision prevention device, a vehicle control method relating to collision prevention, and a vehicle control program. [Background technology]

[0002] Patent Document 1 discloses a collision avoidance device that detects an approaching object and avoids a collision with the detected approaching object. The collision avoidance device includes a radar that detects an object located behind the vehicle and detects the distance to the detected object, a plurality of ultrasonic sensors that detect an object located behind the vehicle and detects the distance to the detected object, and a plurality of ultrasonic sensors each having a different detection area. The collision avoidance device detects an approaching object approaching the vehicle from among objects detected by the radar, and when an object is detected at multiple positions by the multiple ultrasonic sensors, it estimates the presence of an obstruction across the multiple positions where the object is detected, and when the distance to the approaching object is greater than the distance to the obstruction by a predetermined value or more, it restricts or prohibits the execution of driving assistance to avoid a collision with the detected approaching object.

[0003] Patent Document 2 discloses an alarm device that is mounted on a vehicle and that notifies the driver of the vehicle of an alarm candidate (same as the approaching object) when the vehicle reverses, and suppresses notification of an alarm candidate present in a shielded area, which is an area on the opposite side of the shielded boundary as seen from the vehicle. The alarm device projects a search wave to the right rear and left rear of the vehicle, which is a search range, and acquires target information including the position of at least one target detected within the search range from a radar module that detects reflected waves resulting from the search wave. Using the acquired target information, the alarm device determines whether each of the at least one target identified from the target information is an alarm candidate that requires notification to the driver of the vehicle, and notifies the driver of the alarm candidate when the vehicle reverses. The warning device also uses the reflection point information acquired by receiving the reflected wave to extract a stationary reflection point, which is a stationary reflection point, from at least one reflection point identified from the reflection point information, calculates an approximation line by performing a robust estimation of the position of the stationary reflection point identified from the reflection point information, and sets the approximation line as an occlusion boundary (the same as the above-mentioned occlusion). The warning device suppresses notification by the notification unit regarding warning candidates existing in an occlusion region, which is a region on the opposite side of the occlusion boundary as viewed from the vehicle itself. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2017-13756 A [Patent Document 2] JP 2020-154786 A Summary of the Invention [Problem to be solved by the invention]

[0005] The present disclosure provides a vehicle control device, a vehicle, a vehicle control method, and a vehicle control program that can accurately determine whether or not a detected approaching object should be excluded from collision determination. [Means for solving the problem]

[0006] The present disclosure relates to a vehicle control device mounted on a vehicle, the vehicle control device including an acquisition unit that acquires detection information that detects an obstacle around the vehicle, and a sensor control unit that performs a collision determination that evaluates the possibility of a collision with the obstacle, the sensor control unit generates information on an approaching object that is an obstacle approaching the vehicle and information on a detection point cloud that is a collection of detection points that indicate a stationary obstacle based on the detection information, and if the detection point cloud has a shielding effect that shields the vehicle from the approaching object, the approaching object is excluded from the collision determination, and the shielding effect is evaluated using a gap threshold based on a vehicle width. A vehicle control device, wherein the detection information includes detection information detected by a sonar, the sensor control unit does not exclude a detection point detected by the sonar from the collision determination, and the sensor control unit selects a detection point to be evaluated for the shielding effect from the detection points detected by the sonar based on a reception strength of the sonar. A vehicle control device is provided.

[0007] The present disclosure also provides a vehicle collision prevention system including an obstacle detection means for detecting an obstacle around a vehicle and outputting detection information, and a sensor control unit for performing a collision determination for evaluating a possibility of a collision with the obstacle, the sensor control unit generating information on an approaching object, which is an obstacle approaching the vehicle, and information on a detection point cloud, which is a collection of detection points indicating an obstacle that does not move, based on the detection information, and excluding the approaching object from the collision determination when the detection point cloud has a shielding effect of shielding the vehicle from the approaching object, and evaluating the shielding effect using a gap threshold based on a vehicle width. A vehicle, wherein the detection information includes detection information detected by a sonar, the sensor control unit does not exclude a detection point detected by the sonar from the collision determination, and the sensor control unit selects a detection point to be evaluated for the shielding effect from the detection points detected by the sonar based on a reception strength of the sonar. Provide the vehicle.

[0008] The present disclosure also provides a vehicle control method executed by one or more computers mounted on a vehicle, which obtains detection information detecting an obstacle around the vehicle, generates information on an approaching object that is an obstacle approaching the vehicle and information on a detection point cloud that is a collection of detection points indicating a stationary obstacle based on the detection information, evaluates a shielding effect of the detection point cloud shielding the vehicle from the approaching object based on the information on the approaching object and the information on the detection point cloud, and excludes the approaching object from targets for collision determination that evaluates a possibility of a collision with the vehicle according to the shielding effect, and evaluates the shielding effect using a gap threshold based on a vehicle width. A vehicle control method, wherein the detection information includes detection information detected by a sonar, and detection points detected by the sonar are not excluded from the collision determination, and detection points to be evaluated for the shielding effect are selected from the detection points detected by the sonar based on a reception intensity of the sonar. A method for controlling a vehicle is provided.

[0009] The present disclosure also provides a vehicle control program executed by one or more computers mounted on a vehicle, the vehicle control program including the steps of: acquiring detection information of an obstacle detected around the vehicle; generating, based on the detection information, information of an approaching object that is an obstacle approaching the vehicle and information of a detection point cloud that is a collection of detection points indicating an obstacle that does not move; evaluating, based on a gap threshold based on a vehicle width, the information of the approaching object, and information of the detection point cloud, a shielding effect of the detection point cloud shielding the vehicle from the approaching object; and excluding the approaching object from targets of collision determination for evaluating a possibility of a collision with the vehicle according to the shielding effect. A vehicle control program, wherein the detection information includes detection information detected by a sonar, and detection points detected by the sonar are not excluded from the collision determination, and detection points to be evaluated for the shielding effect are selected from the detection points detected by the sonar based on a reception intensity of the sonar. Provides a vehicle control program. Effect of the Invention

[0010] According to the present disclosure, it is possible to determine with high accuracy whether or not a detected approaching object should be excluded from collision determination. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram showing an example of an internal configuration of a host vehicle according to a first embodiment; [Diagram 2] FIG. 1 is a diagram showing an example of the internal configuration of a sonar and a radar according to a first embodiment; [Diagram 3] A diagram explaining sonar detection and judgment processing [Figure 4] FIG. 1 is a diagram for explaining an example of the arrangement and detection range of a sonar of a host vehicle according to the first embodiment; [Diagram 5] FIG. 1 is a diagram for explaining an example of the arrangement and scanning range of a radar of a host vehicle according to a first embodiment; [Figure 6] A diagram explaining the estimated ghost position [Figure 7] FIG. 1 is a diagram illustrating a process of evaluating ghost likelihood over time. [Figure 8] FIG. 1 is a diagram illustrating a process of evaluating ghost likelihood over time. [Figure 9] A diagram explaining the ghost likelihood weighting process. [Figure 10] A diagram explaining the method for evaluating safety based on the detection direction of radar and sonar [Figure 11] How to assess the safety of gaps [Figure 12] FIG. 1 is a diagram for explaining a method for evaluating the safety of obstacles in the traveling direction of a vehicle. [Figure 13] A diagram explaining the method for determining the detection point group. [Figure 14] A diagram explaining a method for evaluating safety when setting a target parking position. [Figure 15] FIG. 13 is a diagram for explaining the automatic setting process of the parking target position. [Figure 16] FIG. 1 is a diagram for explaining a method for evaluating ghost likelihood and safety when a parking target position is set. [Figure 17] FIG. 1 is a diagram for explaining a method for evaluating ghost likelihood and safety when a parking target position is set. [Figure 18] FIG. 1 is a diagram for explaining a method for evaluating ghost likelihood and safety when a parking target position is set. [Figure 19] A diagram explaining the relationship between the detection of detected objects and approaching objects and the evaluation of ghost likelihood and safety level. [Figure 20] A flowchart showing an example of an operation procedure of a host vehicle according to the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] (Background to this disclosure) The collision avoidance device shown in Patent Document 1, when obstacles are detected at multiple positions by multiple ultrasonic sensors, estimates that there is an obstruction that includes the obstacles at the multiple positions as part of the obstruction. Specifically, if there are two positions where an object is detected, it estimates that there is an obstruction such as a guardrail that passes through the two points. The collision avoidance device detects an approaching object approaching the vehicle from among the objects detected by the radar, and when the distance to the approaching object is greater than the distance to the obstruction by a predetermined value or more, it limits or prohibits the execution of driving assistance that executes collision avoidance with the detected approaching object. When the approaching object detected by the radar is a mirror ghost, that is, a virtual image, caused by the reflection of radar waves by an obstruction such as a guardrail, the distance to the mirror ghost is twice the distance to the obstruction, so that the control to prohibit the above collision avoidance may be appropriate. However, the obstruction estimated by the above-mentioned method may not actually exist as a continuous obstruction that other vehicles cannot pass through. For example, there may be an isolated object at each position where an object is detected, and there may be a space between the objects through which other vehicles can pass. In addition, mirror ghosts are likely to occur when there is a continuous obstruction such as a guardrail that reflects radar waves, so conversely, if there is no continuous obstruction such as a guardrail, it may not be a mirror ghost (i.e., the approaching object may be a real object that may collide with the vehicle). Therefore, it is desirable for the collision avoidance device to determine whether or not the detected object is a mirror ghost (hereinafter, referred to as "ghost") and to determine whether or not driving assistance is required depending on the result of the determination.

[0013] Moreover, the warning device shown in Patent Document 2 calculates an approximation line indicating the position of an obstruction (i.e., an obstruction boundary) based on the position of a stationary reflection point indicating a stationary target acquired by a radar. The warning device calculates an obstruction reliability indicating the likelihood that a detected obstruction exists based on the calculated approximation line, and when it is determined that the calculated obstruction reliability is equal to or greater than an obstruction threshold, it determines that an obstruction exists and suppresses notification by the notification unit regarding an alarm candidate existing in an obstruction area, which is an area on the opposite side of the obstruction boundary as viewed from the vehicle. Specifically, in calculating the obstruction reliability, the warning device sets divided areas by dividing the xy plane along the y-axis direction, and calculates the obstruction reliability for each divided area. However, since the divided areas are set every 12 m in the vehicle width direction, there are cases where the obstruction reliability of a divided area is calculated to be high even though there are parts in the divided areas where there are no stationary reflection points and other vehicles can pass through. In other words, when evaluating the reliability of an obstruction, the spacing between stationary reflection points should be evaluated based on the vehicle width. Even if the detected stationary targets are distributed linearly and the presence of an obstruction such as a guardrail is estimated, if there is a gap in the obstruction that is larger than the width of one vehicle (a gap that is wider than the vehicle width), it is desirable for the warning device not to exclude the extracted vehicle located on the other side of the obstruction from the target of the warning.

[0014] Hereinafter, with reference to the drawings as appropriate, each embodiment specifically disclosing a vehicle control device, a vehicle, a vehicle control method, and a vehicle control program according to the present disclosure will be described in detail. However, more detailed explanation than necessary may be omitted. For example, detailed explanation of already well-known matters and duplicate explanation of substantially the same configuration may be omitted. This is to avoid the following explanation becoming unnecessarily redundant and to facilitate understanding by those skilled in the art. Note that the attached drawings and the following explanation are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0015] First, the internal configuration of the host vehicle C1 according to the first embodiment will be described with reference to Fig. 1, Fig. 2, and Fig. 3. Fig. 1 is a diagram showing an example of the internal configuration of the host vehicle C1 according to the first embodiment. Fig. 2 is a diagram showing an example of the internal configuration of the sonar 12 and the radar 13 according to the first embodiment. Fig. 3 is a diagram explaining the detection and determination process of the sonar 12.

[0016] The host vehicle C1 according to the first embodiment is not limited to a vehicle manually driven by a driver, but may be an autonomous vehicle. The host vehicle C1 moves backward or forward autonomously to park in a set parking position or travel toward a set destination.

[0017] The vehicle C1 includes a vehicle control device 20, which is an example of a computer. The vehicle control device 20 includes a mechanical sensor 10, at least one camera 11, twelve sonars 12, three radars 13, a memory 14, an HMI (Human Machine Interface) 15, a sensor control unit 16, a vehicle control unit 17, a navigation system (GPS: Global Positioning System) 18, and an in-vehicle LAN (Local Area Network) 19.

[0018] The in-vehicle LAN 19 is connected to each part mounted on the vehicle C1 so that data can be transmitted and received between them. The parts referred to here are the mechanical sensor 10, at least one camera 11, twelve sonars 12, three radars 13, memory 14, HMI 15, sensor control unit 16, vehicle control unit 17, and navigation system 18. In FIG. 1, sensors such as the mechanical sensor 10, the camera 11, the sonar 12, and the radar 13 are also included in the elements of the vehicle control device 20, but the grouping of the block diagram in the vehicle is arbitrary, so they may be configured in a different grouping. For example, the vehicle control device 20 may be configured to include the memory 14, the sensor control unit 16, and the vehicle control unit 17, and the sensors such as the mechanical sensor 10, the camera 11, the sonar 12, and the radar 13 may be connected to the vehicle control device 20 via the in-vehicle LAN 19, and the vehicle control device 20 may process information obtained by the sensors to control the vehicle.

[0019] The mechanical sensor 10 is a sensor that measures, for example, the steering angle, gear position, speed information, etc. of the host vehicle C1. The mechanical sensor 10 outputs the measurement results to the sensor control unit 16 via the in-vehicle LAN 19.

[0020] The camera 11 has a solid-state imaging element (image sensor) such as a CCD (Charged-Coupled Device) or a CMOS (Complementary Metal-Oxide-Semiconductor), forms an image of light from a subject, converts the formed optical image into an electrical signal, and outputs a video signal. The camera 11 outputs a video signal output from a captured image (captured video) to the sensor control unit 16. The host vehicle C1 is equipped with at least two cameras 11 and captures images in the front, rear, left and right directions of the host vehicle C1.

[0021] The twelve sonars 12, which are an example of an acquisition unit and an obstacle detection means, are controlled by the sensor control unit 16 to detect objects located in the front, rear, left and right of the host vehicle C1. The twelve sonars 12 output detection information, which associates data including distance information to the detected object with identification information that can identify the sonar, to the sensor control unit 16. The host vehicle C1 is equipped with the twelve sonars 12 on the bumper as shown in FIG. 4, and is arranged so as to be able to detect objects located in the front, rear, left and right of the host vehicle C1. The sonars 12 include a controller 12A, a drive circuit 12B, a piezoelectric element 12C, and a receiving circuit 12D.

[0022] Based on a control command output from the sensor control unit 16, the controller 12A causes the drive circuit 12B to generate an AC voltage, applies the AC voltage to the piezoelectric element 12C, and causes the piezoelectric element 12C to emit ultrasonic waves. The piezoelectric element 12C receives the reflected wave of the emitted ultrasonic waves, converts it into an AC voltage, and outputs the AC voltage to the receiving circuit 12D. The receiving circuit 12D amplifies and detects the AC voltage output from the piezoelectric element 12C, and outputs the result to the controller 12A. Here, the ultrasonic waves emitted by the piezoelectric element 12C are pulsed ultrasonic waves, and the distance to the object is determined based on the time it takes for the reflected wave (i.e., echo waveform) reflected by an object such as a road surface or an obstacle to be received (detected) by the piezoelectric element 12C. The controller 12A outputs information on the determined distance to the sensor control unit 16.

[0023] The three radars 13, which are an example of an acquisition unit and an obstacle detection unit, are controlled by the controller 13A based on a control command output from the sensor control unit 16. The radar 13 has an array antenna element group 13C arranged in a two-dimensional lattice pattern, and under the control of the controller 13A, the drive circuit 13B applies high-frequency waves to these antenna elements, the phase of which is shifted according to the position on the lattice. The array antenna element group 13C converts the applied high-frequency waves into radar waves and transmits radar waves with directivity in a specific direction according to the phase difference. The drive circuit 13B changes the phase difference between the antenna elements as a function of time, so that the radar 13 periodically swings (scans or sweeps) the transmission direction of the radar wave. When the reflected wave generated by the radar wave being reflected by an object is received by the antenna, the direction of the object that reflected the radar wave is the direction in which the directional radar wave was transmitted, and since the transmission direction of the radar wave is a function of time, the direction of the object can be identified by the reception time of the reflected wave. The antenna for receiving the radar wave reflected by the object may be the array antenna element group 13C or another antenna (not shown). If the transmitted high frequency is appropriately modulated, the transmitting antenna can be used as a receiving antenna as well. When the array antenna element group 13C is used as a receiving antenna, it can be made directional so that it can selectively receive the radar wave in the transmitting direction, so that it is possible to prevent the occurrence of ghosts caused by radio waves coming from a direction other than the transmitting direction of the radar wave (i.e., ghosts whose paths cannot be identified, unlike ghosts whose radio wave paths can be identified such as mirror ghosts). If the transmitted high frequency is FM modulated, the distance to the object that reflected the radar wave can be detected as the difference between the frequency of the received wave and the frequency of the transmitted wave at that time. If the object that reflected the radar wave is approaching, the frequency of the received wave increases due to the Doppler effect, so the approaching speed can be detected by detecting this increment in frequency. The detection of the distance and the approaching speed is performed by the receiving circuit 13D, and the detection results are output to the controller 13A.

[0024] The controller 13A of the three radars 13 outputs detection information, which is data on the direction, distance, and approach speed of a detected object, plus identification information for identifying the radar, to the sensor control unit 16. The vehicle C1 is equipped with three radars 13 as shown in Fig. 5 (left), and is arranged so as to be able to detect objects approaching from the front, right rear, and left rear of the vehicle C1. The radar 13 includes a controller 13A, a drive circuit 13B, an array antenna element group 13C, and a receiving circuit 13D.

[0025] The memory 14 has, for example, a RAM (Random Access Memory) as a work memory used when executing the processing of the sensor control unit 16 and the vehicle control unit 17, and a ROM (Read Only Memory) that stores a program that defines the processing of the sensor control unit 16 and the vehicle control unit 17. The RAM temporarily stores data generated or acquired by the sensor control unit 16 and the vehicle control unit 17. The ROM has a program that defines the processing of the sensor control unit 16 and the vehicle control unit 17 written therein. The memory 14 may include a non-volatile and rewritable magnetic recording device or an electrically rewritable ROM, for example, an EEPROM (Electrically Erasable and Programmable Read-Only Memory) or a flash memory. The position and range of a fixed obstacle such as a guardrail may be stored in these non-volatile memories. Since the information stored in the non-volatile memory is retained even when the vehicle is parked and all power is turned off, for example, when the vehicle leaves the garage, the position information of the obstacle detected when the vehicle is parked can be used.

[0026] The HMI 15 includes input / output devices such as a display, a touch panel, switches (buttons), and a speaker. The touch panel is integrated and mounted on the surface of the display. The switches (buttons) are not limited to mechanical ones, and may function as switches by detecting an operation on a button displayed on the display with the touch panel. The HMI 15 can accept operations by a passenger of the vehicle C1, converts the operation contents accepted by the input devices such as the touch panel and switches (buttons) into an electric signal, and outputs it to the vehicle control unit 17. The HMI 15 also outputs advance notice information for advancing the execution of emergency braking output from the vehicle control unit 17, warning information for notifying the execution of deceleration control, and the like to output devices such as a display and a speaker.

[0027] The sensor control unit 16 is configured using, for example, a CPU (Central Processing Unit) or an FPGA (Field Programmable Gate Array), and performs various processes and controls in cooperation with the memory 14. Specifically, the sensor control unit 16 refers to a program stored in the memory 14 and executes the program to realize various functions. The number of CPUs or FPGAs is not limited to one, and multiple CPUs or FPGAs may be implemented to enable multiple programs to be executed simultaneously. In addition, the housings of sensors such as the camera 11, the sonar 12, and the radar 13 may incorporate a CPU or FPGA that executes some of the functions of the sensor control unit 16. FIG. 1 is a block diagram illustrating the grouping of functions, and does not specify the arrangement within the vehicle.

[0028] The sensor control unit 16 executes transmission control of ultrasonic waves by the sonar 12 and radio waves by the radar 13. The sensor control unit 16 executes control for transmitting ultrasonic waves and radio waves to the sonar 12 and the radar 13, and evaluates the ghost likelihood or the safety level based on the detection information detected by the sonar 12 and the radar 13. The sensor control unit 16 executes a collision determination between the host vehicle C1 and an approaching object based on the evaluated ghost likelihood or safety level, and determines whether emergency braking is necessary. Note that the evaluation of the ghost likelihood and safety level and the collision determination do not evaluate only the instantaneous numerical values at that time, but refer to the time series of the numerical values for processing. The memory 14 stores the time series of the detection information, the ghost likelihood, or the safety level, and the sensor control unit 16 refers to and processes the time series of the data (numerical values) stored in the memory 14 to perform evaluation and determination.

[0029] The host vehicle C1 is equipped with a total of 12 sonars, 4 on each of the front and rear bumpers and 2 on each side. The sonar 12 transmits ultrasonic waves from the piezoelectric element 12C and receives the reflected waves reflected by an object around the vehicle with the same piezoelectric element 12C. The reception circuit 12D generates waveform data of the reflected waves based on the temporal change in the intensity of the reflected waves. The controller 12A converts the time until the rise of the waveform of the generated reflected wave (that is, the echo waveform) into a distance, and calculates the distance between the sonar 12 and the object. The sensor control unit 16 specifies the relative position (relative coordinates) of the object with respect to the vehicle body based on the distances between each sonar 12 and the object calculated by the plurality of sonars according to the principle of trilateration. The position of the object whose position has been specified will hereinafter be referred to as the detection point.

[0030] The sonar 12 receives not only the reflected waves (ultrasound waves) from objects but also the reflected waves from the road surface. The sonar 12 detects obstacles and other objects based on a first threshold for excluding the reflected waves from the road surface from the received reflected waves. Note that the first threshold is set to be lower as the time from when the reflected wave is transmitted to when it is received becomes longer (i.e., the farther the distance from the sonar 12 is), because ultrasound waves attenuate rapidly in the air. By excluding weak reflected waves from relatively small objects, as well as the reflected waves from the road surface, are excluded from detection by this first threshold.

[0031] Furthermore, the sensor control unit 16 judges whether the detected object is an object capable of shielding an approaching object approaching the vehicle C1 based on the intensity of the detected reflected wave. The sensor control unit 16 judges whether the intensity of the detected reflected wave is equal to or greater than a second threshold value that is greater than the first threshold value, and when it is judged that the intensity of the reflected wave is equal to or greater than the second threshold value, it judges that the detected object is a candidate shield. That is, among the detection points, an object having a high intensity of the reflected wave is a candidate shield. If an object is capable of shielding an approaching object, it should have a corresponding size, and if it has a corresponding size, it is expected that the intensity of the reflected wave will also be correspondingly large. Note that, even if the size of an object is the same, the farther the object is, the weaker the reflected wave will be. Therefore, the second threshold value is set to be lower as the time from the time when the reflected wave is transmitted to the time when it is received is longer (i.e., the farther the distance from the sonar 12 is), similar to the first threshold value, and a value larger than the first threshold value is set at the same time (or the same detection distance). When multiple sonars detect multiple potential shields and the coordinates of the multiple potential shields indicate that the potential shields are arranged in a line, the sensor control unit 16 estimates that a shield is present. The coordinates of the multiple potential shields may be detected simultaneously or may be acquired in a time series manner by repeating detection. It is not essential to select potential shields from detection points based on the intensity of the reflected waves. For example, if a threshold value that can detect even relatively small obstacles is not used, but only a threshold value corresponding to a reflected wave from an object with a reasonable size that can shield an approaching object is used for detection, all detection points are also potential shields, so selection is not necessary. Therefore, hereafter, potential shields will not be distinguished from detection points, and all will be called detection points. The likelihood of a shield may be determined based on the positional relationship of the detection points. The evaluation of the positional relationship of the detection points will be explained later.

[0032] The radar 13 mounted on the vehicle C1 periodically scans the transmission direction while transmitting radio waves and simultaneously receiving the reflected waves. Any object around the vehicle irradiated with the radio waves is a reflector that returns a reflected wave, but the radar 13 extracts a reflector approaching the vehicle C1 from among the reflectors as an approaching object and outputs detection information of the approaching object to the sensor control unit 16.

[0033] Specifically, since the radio waves transmitted by the radar 13 propagate at the speed of light, the transmission direction of the radio waves at the time of receiving the reflected waves is the same as the direction of the reflector that reflected the radio waves. Therefore, the direction of the reflector can be identified as the transmission direction of the radio waves. In addition, the distance L between the radar 13 and the reflector can be calculated based on the time difference between the transmission time of the radio waves and the reception time of the reflected waves of the radio waves. The radar 13 identifies the coordinates of the reflector based on the calculated distance L and the transmission direction of the radio waves. In addition, if the reflector is an approaching object, the radio waves reflected by the reflector and received by the radar 13 undergo a Doppler shift in which the frequency of the reflected wave becomes higher than the frequency of the transmitted wave. Since the difference in frequency between the transmitted wave and the reflected wave is proportional to the approaching speed of the reflector approaching the vehicle C1, the radar 13 calculates the approaching speed of the reflector based on the frequency difference between the frequency of the reflected wave and the frequency of the transmitted wave, and outputs detection information including the approaching speed and coordinates of the reflector to the sensor control unit 16.

[0034] The above-mentioned Doppler shift occurs when the vehicle C1 approaches the reflector even if the reflector is a stationary object. Since the detection information of the approaching object received by the sensor control unit 16 includes such detection information of a stationary object, the sensor control unit 16 cancels the approach speed of the reflector calculated based on the speed information of the vehicle C1 output from the mechanical sensor 10 and the direction of the reflector and the approach speed due to the speed of the vehicle C1. If the speed after the cancellation is approximately zero, the detected reflector is determined to be a stationary object (i.e., not an approaching object) and is excluded from the target of the collision determination process described below. In this way, the sensor control unit 16 can distinguish whether the detected reflector is a stationary object or an approaching object.

[0035] Radio waves may be reflected by a reflector such as a guardrail, the reflected waves may be reflected by the vehicle C1, and the reflected waves from the vehicle C1 may be reflected again by the guardrail and received by the radar. This phenomenon is called a mirror ghost, or simply a ghost, because the guardrail acts like a mirror surface on the radio waves and detects a vehicle on the other side of the guardrail. When the vehicle C1 approaches the guardrail, the ghost is also observed to be approaching the guardrail at the same speed, so the approach speed of the ghost to the vehicle C1 is twice the approach speed of the vehicle C1 to the guardrail. The guardrail can also be detected by sonar, so if the position of the reflector is estimated from the detection information of the sonar, and the position where the ghost can be detected calculated from the position of the reflector is roughly the same as the position of the approaching object detected by the radar, it can be evaluated that the estimation that the detected approaching object is a ghost is likely (the ghost likelihood is high). The sensor control unit 16 judges whether the approaching object detected by the radar 13 is a ghost or not based on the evaluated ghost likelihood. If the approaching object is a ghost, it will not collide with the host vehicle C1, so the sensor control unit 16 evaluates the situation as being safe. In addition, if a continuous object such as a guardrail is detected by sonar, if the approaching object is located on the other side of the continuous object, it is expected that the continuous object will act as a shield and block the approach of the approaching object, so the situation is evaluated as being safe. The sensor control unit 16 determines whether or not the host vehicle C1 will collide with the approaching object based on the evaluated safety level (collision determination process). If the sensor control unit 16 determines that the host vehicle C1 will collide with the approaching object within a predetermined time as a result of the collision determination, it causes the vehicle control unit 17 to execute emergency braking or deceleration control.

[0036] The vehicle control unit 17 executes various controls (e.g., steering angle adjustment, forward movement, reverse movement, emergency braking, deceleration control, etc.) on the motion of the host vehicle C1 based on the control information output from the sensor control unit 16. When the vehicle control unit 17 executes emergency braking based on the control information output from the sensor control unit 16, it executes emergency braking after causing the HMI 15 to output a notice or warning that emergency braking will be executed.

[0037] The navigation system 18 is a navigation system incorporating a magnetic compass, an acceleration sensor, and a GPS. The navigation system 18 receives satellite positioning signals transmitted from an artificial satellite (not shown), and calculates the position information of the vehicle C1 based on the received satellite positioning signals. The navigation system 18 can also determine the position of the vehicle C1 even when it cannot receive satellite positioning signals by continually determining the amount of movement and the direction of movement of the vehicle C1 using the built-in magnetic compass and acceleration sensor, and speed information obtained from the vehicle control unit 17. The navigation system 18 stores map information, and generates an image showing the position of the vehicle C1 on the map, and outputs it to the HMI 15.

[0038] Here, the detection ranges of the 12 sonars 12 equipped in the host vehicle C1 will be described with reference to Fig. 4. Fig. 4 is a diagram for explaining an example of the arrangement and detection range of the sonars 12 of the host vehicle C1 according to the embodiment 1. Note that the example of the arrangement and the detection range of the 12 sonars shown in Fig. 4 are merely examples, and are not limited thereto.

[0039] The host vehicle C1 is equipped with four sonars FRC, FR, FL, FLC on the front bumper of the host vehicle C1, two side sonars (sonars FRS, BRS) on the right side of the host vehicle C1, two side sonars (sonars FLS, BLS) on the left side of the host vehicle C1, and four sonars BRC, BR, BL, BLC on the rear bumper of the host vehicle C1.

[0040] Sonar FRC detects objects within a detection range FRCX. Sonar FR detects objects within a detection range FRX. Sonar FL detects objects within a detection range FLX. Sonar FLC detects objects within a detection range FLCX.

[0041] The four sonars FRC, FR, FL, and FLC mounted on the front bumper are a forward sonar group with a detection range in front of the vehicle C1, and the detection ranges overlap each other as shown in FIG. 4. However, the actual detection range does not have a clear boundary as shown in FIG. 4, and an object can be detected even if it is located a little away from the detection range shown in the figure, if the object is large. For example, an object near the front of the vehicle can be detected by the two sonars FR and FL, but if the object is large, it may also be detected by the sonars FRC and FLC at the corner. The sensor control unit 16 calculates the coordinates (detection point) of the detected object detected in front of the vehicle C1 based on two pieces of detection information output from two of the four sonars FRC, FR, FL, and FLC. A long object such as a guardrail may be detected simultaneously by three or more sonars. Since one coordinate is calculated for a combination of detection information from two sonars, when an object is detected simultaneously by three or more sonars, multiple coordinates (detection points) may be calculated simultaneously. For example, if the reflected waves of the sound waves emitted by sonar FR can be detected by three sonars, sonar FRC, FR, and FL, three coordinates (detection points) can be calculated in a single detection by using a trilateration with the line segment connecting sonar FRC and FR as one side, a trilateration with the line segment connecting sonar FR and FL as one side, and a trilateration with the line segment connecting sonar FRC and FL as one side.

[0042] The detection range of the sonar FRS mounted on the right side (front side) of the vehicle C1 moves from detection range FRSX1 to detection range FRSX2 as the vehicle C1 moves and the position of the sonar FRS moves. The detection range FRSX1 is the detection range of the sonar FRS at the driving position of the vehicle C1 at time t11. The detection range FRSX2 is the detection range of the sonar FRS at the position of the vehicle C1 at time t12, which is one detection interval ahead in time. An object on the side of the vehicle C1 is detected multiple times as the vehicle C1 drives along the side, and multiple pieces of distance information are obtained starting from multiple sonar positions. The sensor control unit 16 calculates the coordinates of the detected object on the right side (front side) of the vehicle C1 by trilateration based on multiple pieces of detection information output in time series from the sonar FRS. For example, if the distance to an object on the right side is detected three times, at least two detection points are obtained.

[0043] The sonar BRS mounted on the right side (near the rear) of the vehicle C1 has a detection range on the right side (near the rear) of the vehicle C1, but when there is a large object such as a guardrail on the side, the sonar BRS may detect the reflected wave of the sound wave emitted by the sonar FRS, and in the reverse direction, the sonar FRS may detect the reflected wave of the sound wave of the sonar BRS. The sensor control unit 16 basically calculates the coordinates of the detected object on the side based on multiple detection information output in time series from the sonar BRS, but when the sonar FRS also detects a reflected wave, it additionally calculates the coordinates of the detected object by trilateration with the line segment connecting the sonar FRS and the sonar BRS as one side. The sonar FRS and the sonar BRS can be called a right sonar group with a detection range on the right side of the vehicle C1.

[0044] The left sonar group (sonars FLS, BLS) is mounted on the left side of the host vehicle C1 and has a detection range to the left side. The detection mechanism of the left sonar group (sonars FRS, BRS) is the same as that of the right sonar group (sonars FRS, BRS), so an explanation will be omitted.

[0045] Sonar BRC detects objects within detection range BRCX. Sonar BR detects objects within detection range BRX. Sonar BL detects objects within detection range BLX. Sonar BLC detects objects within detection range BRCX.

[0046] The four sonars BRC, BR, BL, and BLC mounted on the rear bumper of the vehicle are a rear sonar group whose overall detection range is behind the vehicle. The detection mechanism of the rear sonar group is the same as that of the front sonar group (sonars FRC, FR, FL, and FLC), so an explanation is omitted.

[0047] As described above, the host vehicle C1 according to the first embodiment can detect objects located around the host vehicle C1 using the 12 sonars, but not all directions are detectable, and there are blind spots that cannot be detected by the sonars. For example, when the vehicle moves forward, an object that is out of the detection range of the sonar FRC on the right corner to the right is not detected by the sonar until the host vehicle C1 moves forward and enters the detection range of the sonar FRS on the right side. The same is true between the detection ranges of the sonar BRC and sonar BRS when the vehicle moves backward.

[0048] Next, the three radars LS1, LS2, and LS3 equipped on the vehicle C1 will be described with reference to Fig. 5. Fig. 5 (left) <during travel> is a diagram for explaining an example of the arrangement and scanning range of the radars LS1 to LS3 of the vehicle C1 according to the embodiment 1. Note that the example of the arrangement and scanning range of the three radars shown in Fig. 5 (left) <during travel> is merely an example, and is not limited thereto.

[0049] The radar LS1 is provided in front of the host vehicle C1 and is used for the forward collision prevention function of the host vehicle C1. The radar LS1 detects reflectors in a scanning range L1AR. The radar LS1 has a narrower scanning range L1AR than the scanning ranges L2AR and L3AR of the other radars LS2 and LS3 in order to detect reflectors (e.g., other vehicles) located at a greater distance from the host vehicle (host vehicle C1) more quickly.

[0050] Radar LS2 is provided at the left rear of the host vehicle C1 and detects reflectors within the scanning range L2AR. Radar LS3 is provided at the right rear of the host vehicle C1 and detects reflectors on the right rear side within the scanning range L3AR. Note that radars LS2 and LS3 are used for the blind spot warning function. Here, the blind spot warning function is a function that detects other vehicles (e.g., other vehicle C3) traveling diagonally rearward of the host vehicle C1, which is likely to be a blind spot for the driver during driving, and notifies (warns) the driver of the presence of the detected other vehicle.

[0051] Here, an example of detecting other vehicle C2 executed by the host vehicle C1 during driving shown in FIG. 5 (left), an example of emergency braking, and an example of detecting other vehicle C31 and blind spot warning will be described.

[0052] First, an example of emergency braking will be described. Here, it is assumed that the host vehicle C1 during driving shown in FIG. 5 (left) is moving straight ahead. Radar LS1 of the host vehicle C1 transmits radio waves in front of the host vehicle C1 and receives reflected waves reflected by other vehicle C2 (i.e., a reflector) traveling in front of the host vehicle C1. Radar LS1 outputs detection information (e.g., presence or absence of a detected object, azimuth, distance, approach speed) regarding the received reflected waves to the sensor control unit 16.

[0053] The sensor control unit 16 refers to the detection information regarding the reflected waves output from radar LS1. When there is a detected object, its azimuth hits the traveling direction of the host vehicle C1, and the distance is equal to or less than a predetermined threshold value, a collision determination is made. At this time, when other vehicle C2 is stopped, the approach speed included in the detection information becomes a value corresponding to the vehicle speed of the host vehicle C1. As a result of the collision determination, if there is a possibility of a collision within a predetermined time, the sensor control unit 16 sends a command to HMI15 to notify the driver of a warning. When the driver does not perform avoidance by steering or braking by a braking operation within the predetermined time, it commands the vehicle control unit 17 to execute emergency braking.

[0054] Next, an example of blind spot warning will be described. In FIG. 5 (left), it is assumed that the host vehicle C1 starts steering to the right while traveling. This steering changes the moving direction of the host vehicle C1 to the right diagonally forward. The radar LS3 transmits radio waves to the right rear of the host vehicle C1, and receives the reflected waves reflected by another vehicle C31 (i.e., a reflector) approaching from the right rear of the host vehicle C1. The radar LS3 outputs detection information (e.g., the presence or absence of a detected object, its direction, distance, and approach speed) related to the received reflected waves to the sensor control unit 16.

[0055] The sensor control unit 16 refers to the detection information output from the radar LS3, and if there is a detected object to the right rear, the distance is within a predetermined threshold, and the approaching speed is equal to or greater than a predetermined threshold, the sensor control unit 16 determines that the other vehicle C31 is a risky approaching object and performs a collision determination. In such a case, the sensor control unit 16 estimates the movement trajectory of the other vehicle C31 based on the time series of the detection information, in particular, the time change of the coordinates of the approaching object determined by the direction and distance. The sensor control unit 16 executes a collision determination between the host vehicle C1 and the other vehicle C31 based on the estimated movement trajectory of the other vehicle C31 and the steering information and traveling speed of the host vehicle C1 output from the mechanical sensor 10. If the collision judgment indicates that the vehicle C1 and the other vehicle C31 will collide at position PT00, a warning is output to the HMI 15, and if the driver does not perform an evasive operation by steering within a specified time, a command is issued to the vehicle control unit 17 to intervene in the steering by controlling the steering angle actuator, and return the vehicle C1 to a straight line to avoid the collision.

[0056] An example of detection of another vehicle C32 and emergency braking performed by the host vehicle C1 during backing up as shown in FIG. 5 (right) <during backing up> will be described.

[0057] FIG. 5 (right) <Backing up> shows a state in which the host vehicle C1 is backing out of a parallel parking lot. The radar LS3 transmits radio waves to the right rear of the host vehicle C1 and receives the reflected waves reflected by another vehicle C32 (i.e., a reflector) approaching diagonally rear-right of the host vehicle C1. The radar LS3 outputs detection information (e.g., the transmission time and reception time of the radio waves) regarding the received reflected waves to the sensor control unit 16. The sensor control unit 16 identifies the position of the other vehicle C32 based on the detection information output from the radar LS3. The sensor control unit 16 determines that the other vehicle C32 is an approaching object and sets it as a target for collision determination.

[0058] In such a case, the sensor control unit 16 estimates the movement trajectory of the approaching object based on the approach speed of the approaching object detected by the radar 13. The sensor control unit 16 executes a collision determination process between the host vehicle C1 and the approaching object based on the estimated movement trajectory of the approaching object and the steering information and traveling speed of the host vehicle C1 output from the mechanical sensor 10. In the example shown in FIG. 5 (right), when the sensor control unit 16 determines that the host vehicle C1 and the other vehicle C32 will collide at the position PT01 within a predetermined time, it generates a control command requesting emergency braking and outputs it to the vehicle control unit 17. The vehicle control unit 17 executes emergency braking based on the control command output from the sensor control unit 16, causing the vehicle to avoid a collision.

[0059] As described above, the host vehicle C1 according to the first embodiment can detect approaching objects around the host vehicle C1 using the three radars LS1 to LS3, and perform collision determination processing based on the detection information, thereby automatically avoiding a collision.

[0060] In the following description, in order to make the description easier to understand, illustrations and reference numerals may be omitted for the 12 sonars 12, the detection ranges of the 12 sonars 12, the three radars 13, and the scanning ranges of the three radars 13. In the following description, examples of the determination processes when the host vehicle C1 moves backward will be described, but it goes without saying that the host vehicle C1 may move forward or be parked forward.

[0061] Next, ghosts and ghost likelihood evaluation processing will be described with reference to Fig. 6. Fig. 6 is a diagram for explaining a ghost estimated position. Note that in Fig. 6, the illustration of the two sonars BRS, BRC and the radar LS3 is omitted, and only the detection ranges of each are shown. Here, a ghost is an approaching object that does not actually exist (a virtual image) and is detected as an approaching object as a result of receiving a reflected wave after a radio wave transmitted from a radar being multiple-reflected.

[0062] For example, when the vehicle C1 (own vehicle) is parking backward and there is an obstruction LN such as a wall behind the parking target position, the radio wave transmitted by the radar may be received after being reflected by the obstruction LN. In such a case, even though there is a wall behind the vehicle C1 and there is no way that another vehicle could be approaching from the right rear side of the vehicle C1, the sensor control unit 16 may erroneously detect an approaching object (ghost) approaching the vehicle C1 from behind the wall. Hereinafter, the position where a ghost is erroneously detected will be described with reference to FIG. 6.

[0063] 6 shows a state in which the vehicle C1 is retreating in a direction X1. When two detection points C and D are detected by two sonars BRS and BRC, the sensor control unit 16 estimates that there is an obstruction LN passing through these detection points C and D. Here, when the radio wave transmitted from the radar LS3 (point A) is reflected by point B on the obstruction LN, the obstruction LN acts like a mirror, and a virtual image (ghost) with no substance may be detected at the position of point G.

[0064] Specifically, when the obstruction LN acts like a mirror, the radio waves of the radar LS3 may be received not only through a simple reflection path from transmission to reception, which is point A (transmission position) → point B → point A (reception position), but also through a multiple reflection path from point A (transmission position) → point B → point A → point B → point A (reception position). Here, point B is the point where a perpendicular line is drawn from point A onto the obstruction LN. In such multiple reflections, the path length from transmission to reception is doubled, so that the radar 13 sees the position (point B) where the reflector is detected as point G. Point G is on the line connecting points A and B, and is twice the distance from point A to point B. In other words, point G is the point that is linearly symmetrical to point A with the obstruction LN as the axis of symmetry. The coordinates of point G are likely to be detected when a ghost occurs due to the presence of an obstruction LN, so if an approaching object is detected and its coordinates are close to point G, it can be determined that the likelihood of it being a ghost (ghost likelihood) is high.

[0065] Approaching objects that have a high ghost likelihood and are determined to be ghosts must be excluded from collision judgment targets. The approaching speed of the ghost erroneously detected at point G approaching the vehicle C1 is calculated to be twice the approaching speed of the approaching object detected by the radar 13 at point B. Therefore, if the ghost is not excluded from collision judgment targets by ghost judgment, it may be determined that there is an approaching object at point G that is rapidly approaching the vehicle C1, and the vehicle control unit 17 may be caused to execute emergency braking.

[0066] In addition, when the radar detects point B as a stationary object, based on the fact that the distance to point G is twice the distance to point B, it may be possible to estimate that point G is a ghost. However, when the distance between point A and point B is equal to or less than the lower limit distance of the detectable range of radar 13, radar 13 may not detect point B (the shielding object LN) and may only detect the ghost of point G. In such a case, the sensor control unit 16 cannot estimate that point G is a ghost because the azimuth of point G is the same as that of point B and the distance to point G is twice the distance to point B. In other words, it may or may not be possible to determine that it is a ghost based on the fact that there is a detection of a stationary object at a position where the azimuth is the same and the distance is half. In the present application, a method for evaluating the likelihood of a ghost based on the positional relationship with the shielding object detected by the sonar is disclosed, but it may be evaluated that the likelihood of a ghost is high when there is a detection of a stationary object at a position where the azimuth is the same and the distance is half.

[0067] In Embodiment 1, the sensor control unit 16 performs ghost estimation to estimate whether the approaching object detected by the radar 13 is a ghost based on the positional relationship with the shielding object detected by the sonar, and evaluates the validity of the ghost estimation (hereinafter referred to as "ghost likelihood").

[0068] With reference to FIGS. 7 and 8, the ghost likelihood evaluation process executed by the sensor control unit 16 will be described. FIG. 7 is a diagram for explaining the ghost likelihood evaluation process over time. FIG. 8 is a diagram for explaining the criteria for ghost likelihood evaluation. In FIG. 7, the host vehicle C1 is traveling from the position PS11 in the direction X1A.

[0069] Based on the detection information output by the sonar 12, the sensor control unit 16 estimates the position of the radio wave reflection surface LN0 that may cause a ghost. The reflection surface LN0 shown in FIG. 7 corresponds to the shielding object LN estimated based on the detection points by the sonar shown in FIG. 6.

[0070] Based on the detection points by the sonar, the sensor control unit 16 calculates the coordinates of the position that is symmetric to the position of the radar 13 with respect to the estimated reflection surface LN0 (for example, the position of point G shown in FIG. 6), which is the position where a ghost may occur (hereinafter referred to as the "ghost estimation position").

[0071] Also, based on the detection information output from the radar 13, the sensor control unit 16 calculates the coordinate difference between the coordinates of the approaching object and the coordinates of the ghost estimation position, or the distance between the position of the approaching object and the ghost estimation position, and evaluates the ghost likelihood based on the calculated distance, coordinate difference, or proximity of the distance. In the detection information output by the radar, since the coordinates are specified by the distance and azimuth, the "coordinate difference (difference in coordinates)" mentioned here refers to the difference in distance and the difference in azimuth. Also, the "distance" mentioned here is, for example, the Euclidean distance. Furthermore, the "proximity of the distance" may be an evaluation value (score) obtained by evaluating the coordinate difference based on the detection error of the radar 13 provided in the host vehicle C1, or an evaluation value (score) obtained by evaluating the Euclidean distance. Here, first, an example of evaluating the ghost likelihood based on the Euclidean distance will be described.

[0072] (Evaluation of Ghost Likelihood Based on Euclidean Distance) When evaluating the ghost likelihood based on the Euclidean distance, the sensor control unit 16 shown in Fig. 7 calculates a position that is linearly symmetrical to the position PS11 as a ghost estimated position PS21 with the reflection surface LN0 estimated based on the detection information output from the sonar 12 as an axis of symmetry when the vehicle C1 is located at the position PS11, calculates the position PS31 of the approaching object based on the detection information output from the radar 13, and calculates the distance L11 between the calculated ghost estimated position PS21 and the position PS31 of the approaching object as the ghost estimation error. If the calculated distance L11 (ghost estimation error) is small, the sensor control unit 16 evaluates the ghost likelihood as high, and if the calculated distance L11 (ghost estimation error) is large, the sensor control unit 16 evaluates the ghost likelihood as low. The ghost likelihood is quantified as an evaluation value (score), and if the ghost estimation error (distance L11) is small, a high evaluation value is given, and if the ghost estimation error (distance L11) is small, a low evaluation value is given.

[0073] Similarly, when the vehicle C1 is located at position PS12, the sensor control unit 16 calculates the estimated ghost position PS22 and the position PS32 of the approaching object, and calculates the distance L12 between the calculated estimated ghost position and the position of the approaching object. The sensor control unit 16 evaluates the ghost likelihood based on the calculated distance L12.

[0074] Similarly, when the host vehicle C1 is located at a position PS13, the sensor control unit 16 calculates the distance L13 between the estimated ghost position PS23 and the position PS33 of the approaching object. The sensor control unit 16 evaluates the ghost likelihood based on the calculated distance L13.

[0075] (Evaluation of ghost likelihood based on radar detection error and coordinate difference) Next, the criteria for evaluating the ghost likelihood will be described with reference to Fig. 8. When evaluating the ghost likelihood based on the detection information of the radar 13, the scale for evaluating the ghost likelihood may be the coordinate difference (difference in coordinates). The radar outputs the distance and direction with the radar position as the origin, so the position of the approaching object is specified on a polar coordinate system. Therefore, the ghost estimated position may be converted to the distance and direction with the radar position as the origin, and the difference between the direction of the estimated ghost position and the direction of the approaching object, and the difference between the distance of the estimated ghost position and the distance of the approaching object may be calculated, and the ghost likelihood may be evaluated based on the difference in direction and the difference in distance (i.e., the difference in coordinates). Since the direction and the distance have different dimensions (units), they cannot be evaluated on a common scale, but each may be non-dimensionalized by dividing it by a standard for the direction difference (e.g., the standard error of the direction of the radar 13) and a standard for the distance difference (e.g., the standard error of the distance of the radar 13), and the ghost likelihood evaluation value (score) may be calculated according to the total value (ratio to the standard error) of the non-dimensionalized values ​​(direction difference ÷ standard error of the direction + distance difference ÷ standard error of the distance), or the ghost likelihood may be evaluated based on the magnitude relationship between the direction difference standard (standard error of the direction) and the distance difference standard (standard error of the distance). For example, as shown in Fig. 8, the sensor control unit 16 evaluates the direction difference and distance difference between the position of the vehicle C1 and the detected object in a polar coordinate system with the position of the radar 13 as the origin, based on the detection information output from the radar 13. The sensor control unit 16 determines whether or not an approaching object is located within the standard error range of the radar 13 centered on the estimated ghost position (a sector-shaped area in which the distance difference from the estimated position is within the standard error range of the distance of the radar 13, and the direction difference from the estimated position is within the standard error range of the direction of the radar 13).

[0076] When the position of the approaching object is position PS41, the sensor control unit 16 determines that the approaching object is located within the standard error range L2AR2 and evaluates the ghost likelihood as "100". On the other hand, when the position of the approaching object is position PS42, the sensor control unit 16 determines that the approaching object is not located within the standard error range L2AR2 and calculates the distance L41 between the approaching object and the standard error range L2AR2. The sensor control unit 16 may multiply the calculated distance L41 expressed in meters by 20 and subtract the result from 100, that is, 100-20×distance [m], as the likelihood. In this way, the likelihood decreases as the object moves away from the standard error range L2AR2, and the likelihood becomes 0 when the object is 5 meters away. When the ghost likelihood is a negative value, the sensor control unit 16 may evaluate the ghost likelihood as 0. As another example, the standard error of the distance and the standard error of the direction may be used as a reference to determine the direction evaluation value AR and the distance evaluation value DT by normalizing the direction difference and the distance difference, respectively, and the ghost likelihood may be calculated from the direction evaluation value AR and the distance evaluation value DT. For example, if the direction evaluation value AR = direction difference ÷ standard error of the direction, and the distance evaluation value AD = distance difference ÷ standard error of the distance, and the ghost likelihood = 100-25 × "square root of (the square of the direction evaluation value AR + the square of the distance evaluation value AD)", the ghost likelihood will be 0 when either the direction difference or the distance difference is four times the standard error.

[0077] The sensor control unit 16 repeatedly performs the above-mentioned evaluation of the ghost likelihood based on the detection information output from the sonar 12 and the radar 13. The sensor control unit 16 associates the ghost likelihood with the evaluation time of the ghost likelihood and stores them in the memory 14, and calculates the total, average, or weighted average of the ghost likelihoods stored in the memory 14 for a predetermined number of recent times (e.g., three times) as the ghost likelihood over time. When calculating the weighted average, it is preferable to give a larger weight to the ghost likelihood stored more recently.

[0078] When the sensor control unit 16 determines that the ghost likelihood over time is equal to or greater than the predetermined likelihood threshold, it determines that the approaching object is a ghost, and when the sensor control unit 16 determines that the ghost likelihood over time is not equal to or greater than the predetermined likelihood threshold, it determines that the approaching object is not a ghost. The sensor control unit 16 may compare the ghost likelihood with a predetermined threshold to determine whether or not the object is a ghost. However, since the ghost likelihood may take a high value by coincidence, a more stable determination result can be obtained by comparing the ghost likelihood over time, which is obtained by evaluating the ghost likelihood over time, with the threshold. Note that if the above-mentioned predetermined number of times is too small, the effect of stabilization is small, and if it is too large, the time until a ghost determination is made becomes long. Therefore, it is desirable to set the above-mentioned predetermined number of times to, for example, 3 times or more and less than 6 times. By setting this predetermined number of times to a small number of times (for example, 3 times), the sensor control unit 16 can shorten the time required to determine whether or not the approaching object is a ghost. For example, if the predetermined number of times is three, the sum of the ghost likelihoods of the most recent three times is calculated as the time-dependent ghost likelihood, and the predetermined likelihood threshold is set to "300", when an approaching object is detected three consecutive times within the standard error range L2AR2 of the radar 13, the sensor control unit 16 can determine that the approaching object is a ghost based on the detection information of these three times. As another time-dependent evaluation method, the approaching object may be determined to be a ghost when the ghost likelihood is equal to or greater than a predetermined threshold for a predetermined number of consecutive times (for example, the ghost likelihood is equal to or greater than 90 three consecutive times). The above-mentioned predetermined likelihood threshold may be set to an arbitrary value based on the standard error range of the radar 13, the evaluation method of the ghost likelihood, the set value of the predetermined number of times, etc.

[0079] In this way, the sensor control unit 16 can evaluate the ghost likelihood over time. When the sensor control unit 16 determines that the approaching object is a ghost based on the evaluated ghost likelihood, it excludes the approaching object from the collision judgment target. On the other hand, when the sensor control unit 16 determines that the approaching object is not a ghost based on the evaluated ghost likelihood, it performs a collision judgment targeting the approaching object, and calculates the time margin until collision from the distance, path, and approach speed of the approaching object. Since the time margin until collision (hereinafter simply referred to as time margin) decreases as the distance to the approaching object decreases, a warning is issued when the time margin falls below a first collision judgment threshold, and when the time margin also falls below a second collision judgment threshold, a control command requesting emergency braking is generated and output to the vehicle control unit 17. The vehicle control unit 17 executes emergency braking based on the control command output from the sensor control unit 16. When a warning or emergency braking is issued, if the detected approaching object is a ghost and there is no approaching object, the passenger (user) will recognize that the collision prevention device has malfunctioned, and user satisfaction will decrease. Therefore, if the approaching object is a ghost, it is sufficient to correctly determine that it is a ghost before the time margin falls below the first collision determination threshold. In other words, when an approaching object is detected, it is not necessary to correctly determine that it is a ghost from the beginning, and it is more effective to evaluate the ghost likelihood over time and obtain a stable ghost determination result.

[0080] Next, the weighting process of the ghost likelihood will be described with reference to Fig. 9. Fig. 9 is a diagram for explaining the weighting process of the ghost likelihood.

[0081] 9 indicate the positions of objects detected within the detection ranges BRCX11, BRSX11 of the sonar 12 when the host vehicle was located at C1A (time T1). Detection points CC02, DD02 indicate the positions of objects detected within the detection ranges BRCX12, BRSX12 of the sonar 12 when the host vehicle was located at C1B (time T2).

[0082] The position GG01 is an estimated ghost position estimated by the sensor control unit 16 when the host vehicle was located at C1A. The position GG02 is an estimated ghost position estimated by the sensor control unit 16 when the host vehicle was located at C1B.

[0083] When two detection points CC01, DD01 are detected based on the detection information output from the sonar 12, the sensor control unit 16 generates an approximate straight line LS based on these detection points CC01, DD01. The number of detection points used to generate the approximate straight line LS may be two or more. The sensor control unit 16 may also generate the approximate straight line LS by excluding detection points that are outliers in the generation of the approximate straight line LS (isolated detection points that are far away from other detection points) from among the detection points used to generate the approximate straight line LS.

[0084] The sensor control unit 16 may generate the approximate straight line LS using regression analysis that analyzes the positions (coordinates) of the detection points detected by the sonar 12, or may generate the approximate straight line LS using other methods (e.g., mean square error, correlation coefficient, (co)variance, etc.).

[0085] (Ghost likelihood weighting process) The sensor control unit 16 evaluates the variation of the detection points with respect to the approximate straight line LS, and calculates a reflection surface likelihood indicating the likelihood that a reflection surface exists on the approximate straight line LS. This variation may be a variance. The variance here is a statistic corresponding to the distance between the approximate straight line LS and each detection point, and instead of the variance, the average value of the distance between the approximate straight line LS and each detection point may be used as the variation. If the variation of the detection points is small, the detection points are converged near the approximate straight line LS, so the reflection surface likelihood is evaluated high, and if the variation of the detection points is large, the reflection surface likelihood is evaluated low. If the reflection surface likelihood is high and the existence of a linear reflection surface is likely, it is also likely that the reflection surface is causing the ghost, but conversely, if the reflection surface likelihood is low and the existence of a linear reflection surface is doubtful, it is also doubtful that the reflection surface is causing the ghost. Therefore, the sensor control unit 16 may determine whether the calculated reflecting surface likelihood is equal to or greater than a predetermined reflecting surface likelihood threshold, and when it determines that the reflecting surface likelihood is less than the predetermined reflecting surface likelihood threshold, it may determine that there is no planar object at the position of the approximate straight line LS, and may set the ghost likelihood of an approaching object (not shown) on the other side of the approximate straight line LS to zero. Alternatively, a correction process may be added such that the ghost likelihood becomes lower when the reflecting surface likelihood is low, for example by multiplying the ghost likelihood by the reflecting surface likelihood. Conversely, a correction process may be added such that the ghost likelihood becomes higher when the reflecting surface likelihood is high.

[0086] Also, the ghost likelihood and the reflection surface likelihood may be evaluated according to the difference between the azimuth of the detection point and the azimuth of the approaching object. For example, at time T1 (when the host vehicle was located at C1A), the approaching object was detected at the same position as the ghost estimated position GG01, and there were detection points CC01 and DD01 detected by the sonar in that direction. Assume that at time T2 (when the host vehicle was located at C1B), the approaching object was detected at the same position as the ghost estimated position GG02, and there were detection points CC02 and DD02 detected in that direction. Then, since the position of the approaching object coincides with the ghost estimated position at any time, the ghost likelihood may be set to the highest point. However, the ghost estimation is based on the premise that there is a reflection surface in the direction of the approaching object. Specifically, it is estimated that there is a reflection surface at the intersection of the line connecting the position of the approaching object and the position of the radar and the approximate straight line LS (in this case, points BB01 and BB02). Therefore, when it is unlikely that there is a reflection surface in the direction of the approaching object (the reflection surface likelihood is low), it is advisable to add a correction process to lower the ghost likelihood.

[0087] At time T1, since the direction of the approaching object is the direction of the ghost estimated position GG01 and the direction of the approaching object is between the detection points CC01 and DD01, the azimuth differences between the detection points CC01 and DD01 and the direction of the approaching object are evaluated. Since the detection point DD01 is away from the line connecting the approaching object and the radar, it can be said that the contribution (evaluation value) of the likelihood of a reflecting surface (reflecting surface likelihood) near BB01 on the approximate straight line LS is relatively low. However, since the detection point CC01 is close to the line connecting the approaching object and the radar, it can be said that the contribution (evaluation value) of the likelihood of a reflecting surface (reflecting surface likelihood) near BB01 on the approximate straight line LS is high. At this time, the sum of the evaluation value of the detection point DD01 and the evaluation value of the detection point CC01 may be used as the evaluation value of the reflecting surface likelihood, or the larger one of them may be used as the evaluation value of the reflecting surface likelihood. At time T2 as well, since the direction of the approaching object (position of GG02) is between the detection point CC02 and the detection point DD02, for the detection points CC02 and DD02, the azimuth differences between them and the direction of the approaching object are evaluated. Since the detection point CC02 is close to the line connecting the approaching object and the radar, it can be said that the likelihood of a reflecting surface (reflecting surface likelihood) near BB02 on the approximate straight line LS is higher compared to that at time T1. Thus, based on the fact that there is a detection point at a position (azimuth) close to the direction of the approaching object, that is, the azimuth difference between the direction of the approaching object and the direction of the detection point is small, the reflecting surface likelihood may be evaluated as high, and in response to the high reflecting surface likelihood, the ghost likelihood may be evaluated as high. Or, without using the scale of the reflecting surface likelihood, if there is no detection point in the direction of the approaching object, the ghost likelihood may be evaluated as lower than when there is a detection point in the direction of the approaching object, or if there are fewer detection points in the direction of the approaching object, the ghost likelihood may be evaluated as lower than when there are more detection points in the direction of the approaching object. Also, the reflecting surface likelihood or the ghost likelihood may be evaluated based on the interval between the detection points in the direction of the approaching object. Specifically, for a pair of detection points located at positions sandwiching the straight line connecting the vehicle (or radar) and the approaching object (a line extending in the direction of the approaching object), if they are at a close distance, the reflecting surface likelihood and the ghost likelihood are evaluated as high, and if a pair of detection points located at positions sandwiching the line extending in the direction of the approaching object are far apart, the reflecting surface likelihood and the ghost likelihood may be evaluated as low.

[0088] Further, when the number of detection points near the approximate straight line LS detected by the sonar 12 increases over time, the sensor control unit 16 may evaluate the likelihood of the reflecting surface higher in response to the increase in the number of detection points. For example, if the evaluation value of the variation = (average distance from the detection point to the approximate straight line LS) ÷ (square root of the number of detection points near the approximate straight line LS), then even if the average distance is the same, the evaluation value of the variation is lower for a larger number of detection points, and the likelihood of the reflecting surface is higher. If the likelihood of the reflecting surface is reflected in the likelihood of the ghost, as the number of detection points near the approximate straight line LS increases over time, the sensor control unit 16 evaluates the likelihood of the ghost higher. That is, when the number of detection points used for generating the approximate straight line is large, the likelihood of the ghost is evaluated higher than when the number of detection points used for generating the approximate straight line is small.

[0089] For example, at time T1, the likelihood P11 is calculated from the detection points (CC01 and DD01) detected in the direction of the approaching object. At the next time T2, for the group of detection points in the direction of the approaching object, the likelihood P12 is calculated from twice the number of detection points, which is the combination of the newly detected detection points (CC02 and DD02) and the already detected detection points (CC01 and DD01). Subsequently, at time T3, the likelihood P13 is calculated from three times the number of detection points at time T1. If the detection points are gathered near the approximate straight line LS, as time progresses, the likelihoods P11, P12, and P13 increase in order.

[0090] Further, the sensor control unit 16 may evaluate the likelihoods P11, P12, and P13 calculated over time equally, but may also weight the likelihood of the ghost based on the number of detection points used in the calculation. For example, the likelihood of the ghost over time at time T3 is G3 = W1×P11 + W2×P12 + W3×P13. However, when W1 + W2 + W3 = 1, it is possible to set W1 = W2 = W3, but it is also possible to evaluate the larger number of detection points more heavily and set W1 < W2 < W3. Alternatively, if the number of detection points at time T3 is sufficiently large, the likelihoods P11 and P12 may be ignored, and the ghost determination may be performed only based on the latest likelihood P13.

[0091] Furthermore, the sensor control unit 16 may perform weighting in response to the reflecting surface likelihood, or the variance, or the average distance. In other words, the weighting coefficient may be selected to correspond to the calculated reflecting surface likelihood, variance, or average distance. For example, when the variance or average distance is small, the sensor control unit 16 may estimate that the reliability of the likelihood evaluation value is high and set a larger weighting coefficient, and when the variance or average distance is large, the sensor control unit 16 may estimate that the reliability of the likelihood evaluation value is low and set a smaller weighting coefficient. Also, in the case of the occlusion effect described next, when the reflecting surface likelihood, variance, or average distance is small, the occlusion effect may be highly evaluated. The shielding effect is obtained when the detection points (obstacles) in the direction of the approaching object do not have gaps that exceed the width of the vehicle, so it is not necessarily necessary for the detection points to be arranged in a line; however, if the detection points are arranged in a line, the shielding effect can be obtained efficiently and it can also be said that the existence of a linear obstruction is plausible. Therefore, when the reflective surface likelihood is large and the variance or average distance is small, the shielding effect can be evaluated higher, and when the reflective surface likelihood is small, the shielding effect can be evaluated higher than when the variance or average distance is large.

[0092] So far, a method has been described in which a reflecting surface is estimated from a detection point of a sonar, and the ghost likelihood is evaluated from the reflecting surface likelihood, the position of a ghost estimated from the reflecting surface, and the position of an approaching object. However, this may be combined with another ghost likelihood evaluation method, or may be replaced with another ghost likelihood evaluation method. For example, if the approach speed of an approaching object changes in synchronization with the deceleration of the vehicle, the ghost likelihood may be evaluated as high. This is because a mirror ghost detects a mirror image of the vehicle, and therefore when the vehicle decelerates, the mirror ghost also decelerates. Specifically, the deceleration rate of the vehicle and the deceleration rate of the approaching object are calculated, and the ratio between the deceleration rate of the vehicle and the deceleration rate of the approaching object is further calculated. If the fluctuation range of this ratio continues within a predetermined range for a predetermined time, it may be determined that the object is a ghost, or if the fluctuation range of the ratio within the predetermined time is small, the ghost likelihood may be evaluated as high.

[0093] When an approaching object is detected but cannot be determined to be a ghost, the vehicle may be decelerated or stopped by emergency braking to avoid a collision. When the vehicle decelerates, the ghost likelihood can be evaluated based on the ratio of the deceleration rate of the vehicle during deceleration to the deceleration rate of the approaching object. Therefore, the deceleration rate ratio may be added to the ghost likelihood determination during deceleration, or the ghost likelihood determination may be performed based only on the deceleration rate ratio during deceleration, or the deceleration rate ratio may be added to the ghost likelihood determination regardless of whether the vehicle is decelerating or not. Using the ghost determination based on the deceleration rate ratio has the effect of increasing the probability of making a ghost determination when decelerating.

[0094] Next, the evaluation process of the shielding effect and the safety level will be described with reference to Fig. 10. Fig. 10 is a diagram for explaining the evaluation method of the safety level based on the detection directions of the radar 13 and the sonar 12.

[0095] (Methods for evaluating shielding effectiveness and safety) When there is a detection point (a detection point included in the detection point group CC1 in the example shown in FIG. 10) detected by the sonar 12 on or near a straight line connecting the position of the approaching object P detected by the radar 13 and the position of the vehicle C1B (specifically, the position A12 of the radar 13), the sensor control unit 16 estimates that the detection point group detects a shielding object and that the detected shielding object may shield the approaching object, evaluates that the detection point group has a shielding effect, and evaluates the safety level against the approaching object as high due to the shielding effect of the detection point group. In other words, since the approach of the approaching object is shielded by the shielding object, it is evaluated that there is no risk of the approaching object colliding. Conversely, when there is no detection point in the direction of the approaching object P, it is estimated that there is no shielding object that has a shielding effect capable of shielding the approach of the approaching object, and evaluates the safety level against the approaching object as low. In other words, it is evaluated as high risk.

[0096] Specifically, the sensor control unit 16 calculates the distance between the straight line connecting the approaching object and the vehicle and the detection point, and evaluates the shielding effect based on the calculated distance. The representative point of the vehicle connected to the approaching object by a straight line may be the corner closest to the approaching object, or the position of the radar that detected the approaching object may be the representative point of the vehicle. In the latter case, the direction of the approaching object detected by the radar coincides with the direction of the line connecting the approaching object and the representative point, which is convenient in terms of calculation. For example, the sensor control unit 16 estimates that the smaller the calculated distance is, the more likely the detection point (or a shield located in a range including the detection point) is to shield the approach of the approaching object, and evaluates the shielding effect higher, and estimates that the larger the calculated distance is, the more likely the detection point is to shield the approach of the approaching object, and evaluates the shielding effect lower. Alternatively, the direction difference between the direction of the approaching object relative to the vehicle and the direction of the detection point relative to the vehicle may be calculated as a reference direction, and a detection point with a small direction difference may be estimated to be highly likely to block the approach of the approaching object and the shielding effect may be evaluated high, while a detection point with a large direction difference may be estimated to be less likely to block the approach of the approaching object and the shielding effect may be evaluated low. Also, since the more detection points there are in the direction of the approaching object, the more likely it is that the object corresponding to the detection point will block the approaching object, the shielding effect may be evaluated for each of the detection points included in the detection point group CC1, and the evaluation values ​​of the shielding effects of the individual detection points may be tallied to determine the shielding effect for the approaching object.

[0097] As described above, when an approaching object P is detected and multiple detection points are detected in the direction of the approaching object P, the sensor control unit 16 estimates that there is a high possibility that the blocking object LN indicated by the multiple detection points can block the approaching object from approaching the vehicle C1, and evaluates the blocking effect as high. If the blocking effect of the detection point group in the direction of the approaching object is high, the safety level of the approaching object may be evaluated as high.

[0098] The safety level here is an index that indicates the risk or low possibility of the host vehicle C1 colliding with another vehicle, an obstacle, or the like. The safety level may be evaluated including the ghost likelihood, and if the ghost likelihood is high, the safety level may also be evaluated high. However, since the safety level is different from the ghost likelihood, the ghost likelihood and the safety level may not be correlated. For example, even if the ghost likelihood is low and the approaching object does not look like a ghost, if the object detected as the detection point group CC1 is in a position that shields the host vehicle from the approaching object, the shielding effect is high, so the safety level may be evaluated high. The reflection surface likelihood described above also has an aspect similar to the shielding effect, but the reflection surface likelihood is not the same as the shielding effect. Since the detection point group CC1 only needs to detect an object that shields the host vehicle from the approaching object, the detection point group CC1 does not need to be arranged in a straight line. That is, if the detection point group CC1 detects an object sufficient to shield the vehicle from an approaching object, the shielding effect is evaluated as high even if the object does not form a reflective surface. That is, even if the reflective surface likelihood is low, the shielding effect may be high. However, as described above, when the detection point group CC1 is arranged in a straight line, the shielding effect may be evaluated as high. Specifically, the sensor control unit 16 may calculate the reflective surface likelihood, variance, or average distance as a numerical value indicating linearity based on the distribution of the detection points, and may increase or decrease the shielding effect according to the numerical value indicating linearity. If the detection points are arranged in a straight line, the shielding effect can be efficiently obtained, and it can be said that the estimation that there is a linear shielding object is plausible. Therefore, when the reflective surface likelihood is large, that is, when the variance or average distance to the approximate straight line is small, the shielding effect may be evaluated with a higher value, and when the reflective surface likelihood is small, that is, when the variance or average distance to the approximate straight line is large, the shielding effect may be evaluated higher than when the reflective surface likelihood is small. In this case, due to the high linearity of the detection point cloud, both the occlusion effect and the ghost likelihood are evaluated highly, and as a result, the safety level is evaluated highly.

[0099] When the safety level calculated based on the shielding effect is high, the sensor control unit 16 judges that the approaching object, which is the subject of the safety level evaluation, is unlikely to collide with the vehicle C1, and excludes the approaching object, which is the subject of the safety level evaluation, from the subject of the collision judgment based on the evaluated safety level. In the example of the ghost judgment and collision judgment described above, the safety level may also be evaluated based on the ghost likelihood, and the corresponding approaching object may be excluded from the subject of the collision judgment based on the evaluated safety level. Of course, the approaching object may be excluded from the subject of the collision judgment according to the safety level calculated based on the shielding effect and the ghost likelihood, or the approaching object may be excluded from the subject of the collision judgment based on the shielding effect and the approaching object may be excluded from the subject of the collision judgment based on the ghost likelihood, separately. When the approaching object is excluded from the subject of the collision judgment, emergency braking for the approaching object is not performed. On the other hand, when the calculated safety level is low, the sensor control unit 16 does not exclude the approaching object that is the subject of the safety level evaluation from the subjects of the collision judgment because there is a possibility that the approaching object that is the subject of the safety level evaluation will collide with the vehicle C1, and when it is judged in the collision judgment that the time margin until collision is small or high, it generates a control command requesting emergency braking to the vehicle control unit 17 and outputs it to the vehicle control unit 17. The judgment to request emergency braking is made based on the time margin until collision (hereinafter, may be simply referred to as "time margin"), so even if the safety level is low, emergency braking is not requested immediately. In other words, the collision judgment may be said to be an evaluation of the time margin until collision.

[0100] (Weighting of occlusion effects) Here, since the shielding effect of the shielding object LN changes based on the position of the shielding object LN relative to the moving direction of the approaching object or the moving direction of the host vehicle C1, the shielding effect may be evaluated based on the angle difference (declination angle, azimuth difference) between the moving direction and the direction of the shielding object LN, or a weighting coefficient according to the angle difference may be used to evaluate the shielding effect. Also, as in the previous example, the shielding effect may be evaluated based on the direction of the approaching object. Since the position of the shielding object LN is specified as a set of detection points (detection point group), the shielding effect may be evaluated individually for gaps between detection points belonging to the detection point group, and the shielding effect for the approaching object may be calculated by aggregating the individual evaluation values. However, as a method of tallying up the shielding effect of the gaps between the detection points, a deduction method (subtraction method) is more suitable than an addition method of tallying up the evaluation values ​​of the individual gaps. For example, when the evaluation values ​​of two gaps are 80 and 70 points out of a total of 100 points, the sum is 150 points when calculated by the addition method, whereas the subtraction method evaluates the sum as 20 and 30 points being subtracted, and evaluates the totaled shielding effect as (100-20-30=) 50 points. With this calculation method, the result of calculation remains the same no matter how many other narrow gaps with evaluation values ​​of 100 points exist, so it can be said that this method is more suitable than the addition method, which results in an increase in the total value as the number of gaps increases. In addition, in the calculation, weighting may be performed based on the angle difference (declination angle) of the direction of the gap with respect to a reference direction (for example, the direction of an approaching object). Specifically, the sensor control unit 16 calculates an evaluation value according to the width of the gap at the detection point, and calculates the angle difference (declination angle, azimuth difference) between the direction of the gap at the detection point (the direction of the midpoint of the detection point) and the reference direction (the direction of the approaching object, the moving direction of the approaching object, or the moving direction of the host vehicle C1). After performing this for each pair of detection points, the sensor control unit 16 sets a weighting coefficient corresponding to the calculated angle difference, and performs weighted addition of each evaluation value using the set weighting coefficient. The weighting coefficient is set according to the angle difference (deflection angle, azimuth difference), and the smaller the angle difference (deflection angle, azimuth difference), the larger the coefficient, and the larger the angle difference (deflection angle, azimuth difference), the smaller the coefficient. This can also be said to be that a gap with a small deflection angle from the reference direction is evaluated more heavily than a gap with a large deflection angle from the reference direction.For example, when tallying up using the point deduction method, if the weighting coefficient for a gap with a declination angle of 0 degrees is 1.0 and the evaluation value is 80 points, then 20 points x 1 = 20 points will be deducted, and if the weighting coefficient for a gap with a declination angle of 30 degrees is 0.5 and the evaluation value is 70 points, then 30 points x 0.5 = 15 points will be deducted, and the total shielding effect will be (100 - 20 x 1 - 30 x 0.5 =) 65 points.

[0101] As shown in Fig. 10, there is a gap between the detection range BRSX1 of the sonar BRS provided on the right side of the vehicle C1A and the detection range BRCX of the sonar BRC provided at the rear, and there is a wide gap between the detection point group captured by the right sonar and the detection point group CC1 captured by the sonar BRC. However, the direction of this gap is not the direction of the approaching object P detected by the radar 13, and the direction is significantly different from the traveling direction X3 of the vehicle C1A. Therefore, in the case of Fig. 10, the gap between the detection point group CC1 and the detection point group captured by the right sonar is not given a large weighting coefficient in the evaluation of the shielding effect, and the presence of the detection point group CC1 in the direction of the approaching object P may be heavily evaluated, and the shielding effect may be evaluated as high.

[0102] When the sonar 12 does not detect the direction of the approaching object detected by the radar 13, the sensor control unit 16 may suspend (postpone) the evaluation of the safety level within the range of the time margin until the collision as described above. According to the method of evaluating the shielding effect and ghost likelihood with emphasis on the detection point of the sonar in the direction of the approaching object as described above, even if the safety level is evaluated at the time when there is no detection point of the sonar in the direction of the approaching object, the approaching object cannot be excluded from the target of the collision judgment (it can also be said that the evaluation is useless). Since approaching objects and ghosts that are not dangerous can be excluded from the target of the collision judgment before emergency braking or preparatory braking is performed as a result of the collision judgment, the evaluation of the safety level may be suspended (postponed) until the time to judge the emergency braking or preparatory braking, that is, within the range of the time margin until the collision. Of course, if there is a margin in the processing performance and power consumption and there is an intention to simplify the control flow, it goes without saying that the safety level may be constantly evaluated regardless of the presence or absence of the detection point of the sonar in the direction of the approaching object. At this time, the detection points are not limited to the sonar detection points, and the shielding effect and ghost likelihood may be evaluated by adding the radar detection points. For example, if there is a reflector detected by the radar in the direction of the approaching object P, and the sensor control unit 16 determines that the detected reflector is a stationary object, the coordinates of the detected reflector may be added to the sonar detection point group to evaluate the shielding effect. Since the blind spot between the lateral sonar detection range and the rear sonar detection range is covered by the detection range of the side rear radar, adding the detection points of stationary objects detected by the side rear radar may avoid a situation in which the safety level cannot be effectively evaluated because there are no detection points in the direction of the approaching object. Since the shielding effect of the detection points detected in the direction of the approaching object P is evaluated higher than the shielding effect of the detection points at positions outside the direction of the approaching object P, adding the radar detection points is expected to enable the safety level to be determined at an earlier stage. Furthermore, if a stationary object is in the direction of the approaching object P and the distance to the stationary object is half the distance to the approaching object P, the likelihood that the approaching object P is a ghost (ghost likelihood) is high, so the level of safety can be evaluated as high due to the high ghost likelihood (however, as mentioned above, radar has a dead zone in close range, so it is not guaranteed that a stationary object will be detected even if it is present).When it is impossible to determine that the approaching object is a ghost and there is no detected object at a position where a shielding effect can be expected, the sensor control unit 16 calculates the time margin until the host vehicle C1A collides with the approaching object, and determines whether to execute emergency braking or preparatory braking based on the calculated time margin.

[0103] Here, the calculation process of the time margin will be described. The time margin is obtained by dividing the distance to the approaching object detected by the radar 13 by the approaching speed of the approaching object also detected by the radar 13. Although the position of the radar 13 is different from the position of the host vehicle C1A (center), since the radar 13 is closer to the approaching object than the center of the vehicle body, it is reasonable to calculate the collision time with the host vehicle C1A based on the position of the radar 13. The sensor control unit 16 calculates the time margin by dividing the distance between the radar 13 and the approaching object by the approaching speed of the approaching object. In the following description, the collision determination process using the first time margin corresponding to the case of not decelerating and the second time margin corresponding to the case of decelerating will be described. The first time margin is (distance ÷ approaching speed), and the second time margin is (distance ÷ approaching speed in the case of deceleration).

[0104] When the sensor control unit 16 determines that the first time margin until the host vehicle C1A collides with the approaching object is equal to or greater than the first time threshold (e.g., 5 seconds), the sensor control unit 16 determines that emergency braking is unnecessary and does not output a control command requesting emergency braking. When the sensor control unit 16 determines that the calculated first time margin is less than the first time threshold (e.g., 5 seconds), the sensor control unit 16 calculates a second time margin until the host vehicle C1A collides with the approaching object when the host vehicle C1A is decelerated at a predetermined deceleration rate. The predetermined deceleration rate in this case is a deceleration rate that does not cause discomfort to the occupants and is a deceleration rate that does not feel like sudden braking. Since the vehicle speed during deceleration is not constant, the vehicle speed that is the divisor of the division may be calculated by, for example, calculating the vehicle speed when the vehicle is decelerated at a predetermined vehicle speed for 2 seconds, and using this as the vehicle speed that is the divisor of the division. When the second time margin is equal to or greater than the second time threshold (e.g., 4 seconds), the vehicle is decelerated at a predetermined deceleration rate and does not perform emergency braking. When the vehicle C1A decelerates, the time margin (second time margin) until the collision between the vehicle C1A and the approaching object becomes longer than the first time margin when the vehicle does not decelerate, which has the effect of extending the time until the collision. If the second time margin is less than the second time threshold, deceleration is insufficient, so emergency braking is immediately performed to avoid the collision. The evaluation of the safety level is not limited to the period before the start of preliminary braking or emergency braking, but is performed continuously while preliminary braking or emergency braking is being performed.

[0105] When the first time margin is equal to or greater than the first time threshold, since there is no imminent collision with the approaching object, it corresponds to a time period when neither emergency braking nor deceleration is required. If the detections by the sonar 12 and the radar 13 are repeated during this time period, the number of detection points of the sonar 12 in the direction of the approaching object increases, and the accuracy of the evaluation values of the ghost likelihood and the shielding effect improves. As a result, if a decision is made to exclude the approaching object from the collision determination target because the safety evaluation value exceeds a predetermined threshold before the first time margin falls below the first time threshold, the emergency braking or deceleration targeting the approaching object will not be performed. However, if the state where the sonar 12 does not detect an obstacle in the direction of the approaching object continues and the evaluation value of the ghost likelihood does not increase either, the first time margin will fall below the first time threshold without the safety evaluation value exceeding the predetermined threshold. In that case, while the first vehicle C1A is decelerated at a predetermined deceleration rate for preparatory braking during the period when the second time margin is equal to or greater than the second time threshold, the detections by the sonar 12 and the radar 13, and the evaluations of the ghost likelihood, the shielding effect, and the safety are continued. The control of the preparatory braking is the same as the control of the emergency braking. The sensor control unit 16 outputs a control command requesting braking to the vehicle control unit 17, and the vehicle control unit 17 executes braking (preparatory braking) based on the control command output from the sensor control unit 16.

[0106] The time period during the preliminary braking can be said to be a time period during which the time until the collision is extended by decelerating, and detection information is waited for to raise the evaluation value of the safety level. In addition, when emergency braking is finally performed, the impact (acceleration) felt by the occupants is suppressed by performing the preliminary braking to reduce the vehicle speed. When the number of detection points increases during the preliminary braking, the evaluation value of the ghost likelihood and the shielding effect increases, and the evaluation value of the safety level exceeds a predetermined threshold, the preliminary braking may be terminated at that point. When the calculated second time margin becomes less than the second time threshold (for example, 4 seconds) without the evaluation value of the safety level exceeding the predetermined threshold, the sensor control unit 16 determines that there is a high possibility of a collision between the host vehicle C1A and the approaching object, and outputs a control command requesting emergency braking to the vehicle control unit 17. The vehicle control unit 17 executes emergency braking based on the control command output from the sensor control unit 16. Even during emergency braking, the detection of the sonar 12 and the radar 13, and the evaluation of the ghost likelihood, the shielding effect, and the safety level may be continued. In this case, when the safety level exceeds the threshold during emergency braking, the emergency braking may be stopped at that point. For example, if the vehicle is performing automatic parking at that point, the braking may be stopped and the automatic parking may be performed to the end. Emergency braking often causes discomfort to passengers, but it is even more uncomfortable and inconvenient if automatic parking is interrupted midway. In other words, it can be said that there is a certain degree of effectiveness in stopping emergency braking at the point when it is determined that the approaching object is not dangerous and continuing the movement that was being performed up until that point.

[0107] The only difference between the preliminary braking and the emergency braking is the deceleration rate specified in the control command. The deceleration rate of the preliminary braking is determined based on an acceleration that does not cause discomfort to the occupants, whereas the deceleration rate of the emergency braking is determined based on an acceleration that does not injure the occupants by the seat belt, etc. Therefore, when the emergency braking is performed, the occupants often feel uncomfortable. In addition, the preliminary braking is also a deceleration that the occupants do not expect, so they may feel uncomfortable. In other words, if the approaching object is a ghost or a moving object that is not dangerous on the other side of the obstruction, the expected value of the customer evaluation is the best (no negative evaluation) if the safety exceeds the threshold value and the approaching object is excluded from the collision judgment before the start of the preliminary braking, the second best (small negative evaluation) if it is during the preliminary braking, the bad (medium negative evaluation) if it is during the emergency braking, and the worst (large negative evaluation) if it is after the stop. As mentioned above, if ghost detection is based on the ratio of the deceleration rate of the vehicle itself to that of the approaching object, the probability of a ghost detection during deceleration increases, so if the approaching object is a ghost, the probability of a ghost detection during preparatory braking is high. In other words, by including a preparatory braking stage, the probability of emergency braking being performed due to a ghost, which would result in the worst customer evaluation, can be reduced.

[0108] The first and second time thresholds are set for the purpose of preventing contact when the approaching object is not a ghost but has substance and is not obstructed, so they may be set according to the running speed of the host vehicle C1A and the approaching speed of the approaching object. For example, when the running speed of the host vehicle C1A is 8 km / h and the approaching speed of the approaching object is 6 km / h, the sensor control unit 16 sets the first time threshold to 5 seconds and the second time threshold to 4 seconds. When the running speed of the host vehicle C1A is reduced more than expected due to preliminary braking, the second time threshold may be raised to match the actual speed, and the start of emergency braking may be delayed. Conversely, when the running speed of the host vehicle C1A is not reduced more than expected due to preliminary braking, the deceleration rate is raised for safety and the vehicle is further decelerated. An approaching object that has been determined to be a ghost and excluded from collision determination may be detected by the sonar 12 of the host vehicle C1A. In other words, this is the case when the ghost determination is incorrect. In such a case, the sensor control unit 16 evaluates the detection information of the sonar 12 with a higher priority than the ghost judgment, and therefore, regardless of the result of the ghost judgment, judges whether or not emergency braking is required based on the collision judgment corresponding to the detection information of the sonar 12. In other words, even if an approaching object detected by the radar is excluded from the collision judgment, the sonar detection is treated as a separate matter from the radar detection of an approaching object, so the collision judgment is performed based only on the detection information of the sonar, and emergency braking is executed if necessary.

[0109] Next, as a method for evaluating the degree of safety based on the shielding effect, a method for evaluating the shielding effect based on the vehicle width W0 will be described. Fig. 11 is a diagram for explaining a method for evaluating the shielding effect of gaps between detection points. In the example shown in Fig. 11, the radar 13 of the host vehicle C1 detects an approaching object GG2. The sonar 12 also detects two detection point groups CC2 and DD2 located between the host vehicle C1 and the approaching object GG2. The directions of the two detection point groups CC2 and DD2 do not match the direction of the approaching object GG2, but the range included in the direction is interpreted broadly and the two detection point groups CC2 and DD2 are described as being "generally" in the direction of the approaching object GG2.

[0110] (How to evaluate gaps) The sensor control unit 16 estimates the shielding object LN based on the detection point group CC2 and the detection point group DD2 in the direction of the approaching object GG2. The position of the shielding object LN is, for example, the position of an approximate line generated based on the detection point group CC2 and the detection point group DD2. The sensor control unit 16 extracts a combination of detection points that has the shortest width W1 in the direction along the estimated shielding object LN from among the combinations of the detection points included in the detection point group CC2 and the detection point group DD2. Alternatively, the sensor control unit 16 may extract the detection points that are closest to the line segment connecting the host vehicle C1 and the approaching object GG2 from the detection point group CC2 and the detection point group DD2, respectively, and calculate the distance between the two extracted detection points. In other words, the direction for evaluating the gap width W1 only needs to be "generally" along the direction of the estimated shielding object LN, so if the two detection points on the left and right closest to the line segment connecting the vehicle C1 and the approaching object GG2 are roughly close to the estimated position of the shielding object LN, the linear distance between the two detection points on the left and right may be calculated and used as the gap width W1. In addition, the estimation of the shielding object LN may be omitted in the evaluation of the shielding effect of the gap. For example, when looking from the vehicle C1 in the direction of the approaching object GG2, if two detection points on the left and right close to that direction are at a similar distance from the vehicle C1, it is possible to evaluate the shielding effect (without estimating the shielding object LN) on the assumption that the pair of detection points may shield the approaching object. In that case, a perpendicular line may be drawn from each pair of detection points to the line segment connecting the vehicle C1 and the approaching object GG2, and the sum of the lengths of the two perpendicular lines may be used as the gap width W1.

[0111] The sensor control unit 16 judges whether the calculated gap width W1 is equal to or larger than a gap threshold value based on the vehicle width of one vehicle (for example, 1.8 m). The gap threshold value is based on the vehicle width of the assumed vehicle, but instead of being set to the same value as the vehicle width, a value smaller than the vehicle width may be set as the gap threshold value, or the gap threshold value may be set assuming a small vehicle with a narrower vehicle width than a standard vehicle. For example, since there are small vehicles with a vehicle width of only 1625 mm, the gap threshold value may be set to 1.6 m, or the gap threshold value may be set to 0.5 m assuming a motorcycle. If the calculated gap width W1 is equal to or larger than the gap threshold value, the sensor control unit 16 evaluates the shielding effect of the shielding object LN as low. On the other hand, if the calculated gap width W1 is less than the gap threshold value, the sensor control unit 16 evaluates the shielding effect of the shielding object LN as high. If the gap width W1 is less than the gap threshold value, the possibility of a vehicle passing through the gap is constant at zero, so the evaluation value of the shielding effect is constant at an upper limit value.

[0112] For example, the sensor control unit 16 may assign an upper limit, for example, "100", to the evaluation value of the shielding effect when the calculated gap width W1 is less than 1.8 m, and may evaluate the shielding effect as "40" when the calculated gap width W1 is 1.8 m or more and less than 2.4 m, and may evaluate the shielding effect as "0" when the width W1 is 2.4 m or more. If the gap is less than the vehicle width, it can be assumed that the vehicle will not pass through the gap, and even if there is a gap that exceeds the vehicle width, if the width is close to the vehicle width, it is assumed that the vehicle will decelerate when passing through the gap, so the shielding effect may be evaluated as being higher than when there is no shielding object. If the gap is wide enough to pass through compared to the vehicle width, it is assumed that the vehicle will pass through without decelerating, so the shielding effect may be evaluated to be low to the same extent as when there is no shielding object.

[0113] The relationship between the gap width W1 and the evaluation value of the shielding effect may not be a step function as described above, but may be a function that gradually decreases when the gap width W1 exceeds the vehicle width W0 (for example, 1.8 m). The vehicle width standard is not limited to 1.8 m corresponding to a four-wheeled vehicle, and 0.5 m corresponding to the width of a two-wheeled vehicle may be used as the vehicle width standard. Furthermore, a function may be used in which the first vehicle width standard is 1.8 m, the second vehicle width standard is 0.5 m, and if the gap width W1 is less than 1.8 m and is 0.5 m or more, the evaluation value of the shielding effect increases as the gap width W1 becomes smaller. It is to be noted that this evaluation method is merely an example and is not limited to this.

[0114] From the above, when the sensor control unit 16 determines that the width of the gap is less than the gap threshold based on the vehicle width, it determines that the approaching object is shielded from the host vehicle by the shield LN, and evaluates the shielding effect as high. On the other hand, when the sensor control unit 16 determines that the gap is equal to or greater than the vehicle width W0, it determines that the shielding is insufficient because the approaching object can pass through the gap, and evaluates the shielding effect as low. When there are multiple gaps between the detection points in the direction of the approaching object, the shielding effect may be evaluated for the multiple gaps, and weighting may be performed according to the angle with respect to the traveling direction of the host vehicle and the direction of the approaching object as described above, and the overall shielding effect may be obtained by weighting addition. As a simple method, a gap with a wider width may be selected, and the evaluation value of that gap may be the shielding effect, or a gap closer to the traveling direction of the host vehicle may be selected, and the evaluation value of that gap may be the shielding effect. In general, if the gap between the detection points in the direction of the approaching object or the traveling direction of the host vehicle is narrower than the vehicle width, it may be determined that the host vehicle is safe. This is because even if the approaching object is not a ghost but a real object, the host vehicle is safe if there is an impassable shield. Conversely, if the ghost likelihood exceeds a predetermined threshold and it is certain that the approaching object is a ghost, the vehicle will not collide with the ghost even if there is no occlusion effect, so the vehicle is safe. Also, if the occlusion effect is moderately high and the ghost likelihood is also moderately high, the overall safety level may be evaluated as high. Alternatively, a threshold value for the occlusion effect (occlusion effect threshold) and a threshold value for the ghost likelihood (ghost likelihood threshold) may be set, and it may be determined that the vehicle is safe when either the occlusion effect or the ghost likelihood exceeds the corresponding predetermined threshold, or a threshold value for the safety level (safety level threshold) may be set without setting the occlusion effect threshold or the ghost likelihood threshold, and it may be determined that the vehicle is safe when the safety level calculated from the occlusion effect and the ghost likelihood exceeds the safety level threshold. When it is determined that the vehicle is safe by either method (i.e., the possibility that the approaching object to be determined will collide with the vehicle is sufficiently low), the approaching object to be determined is excluded from the collision determination, so that unnecessary emergency braking or preparatory braking is not performed.

[0115] Next, a method for evaluating the shielding effect in the traveling direction of the host vehicle C1 will be described with reference to Fig. 12. Fig. 12 is a diagram for explaining the traveling direction of the host vehicle C1, the traveling direction of an approaching object, and the position of the shielding object.

[0116] In the example shown in Fig. 12, the host vehicle C1 is moving (backing up) in the traveling direction X4, and the sonar detects a plurality of detection point groups CC3, DD3. The detection point group FF3 is evaluated in the cases of not being detected and being detected at this time. Furthermore, the vehicle C6 detected as an approaching object GG3 is moving (forward) in the traveling direction X5. Furthermore, in Fig. 12, the obstruction LN estimated by the sensor control unit 16 of the host vehicle C1 does not have a gap wider than the vehicle width W0 (see Fig. 11) in the direction in which the approaching object is detected at this time.

[0117] The sensor control unit 16 estimates the shielding object LN based on the multiple detection point groups CC3 and DD3. The shielding object LN is an "estimated shielding object", but is simply referred to as the shielding object LN here. In addition, the sensor control unit 16 evaluates the shielding effect of a gap sandwiching an intersection (evaluation position E1) between the traveling direction X5 of the approaching object GG3 (vehicle C6) detected by the radar 13 and the shielding object LN, in addition to the gap between the detection point group CC3 and the detection point group DD3 sandwiching an evaluation position E2 on the shielding object LN in the direction of the approaching object, as points for evaluating the shielding effect. Since the evaluation of the gap sandwiching the evaluation position E2 has been described above, the evaluation of the gap sandwiching the evaluation position E1 will be described here. Here, the shielding effect is evaluated by setting the intersection of the traveling direction of the approaching object GG3 and the obstruction LN as the evaluation position E1, but the evaluation position E1 may also be the intersection of the traveling direction X4 of the vehicle and the obstruction LN, or the intersection of the traveling direction X5 of the approaching object GG3 and the traveling direction X4 of the vehicle, or the shielding effect may be evaluated at multiple intersections and the lowest evaluation value may be selected.

[0118] Here, first, a description will be given assuming that the detection point group FF3 is detected. The gap evaluation section to be evaluated is a gap between the detection point in the detection point group FF3 that is closest to the evaluation position E1 and the detection point in the detection point group CC3 that is closest to the evaluation position E1. The sensor control unit 16 first obtains the linear distance (width of the gap evaluation section P3) between a pair of detection points sandwiching the evaluation position E1. When the gap evaluation section is regarded as a gap in the shielding object LN and the width is evaluated, if the azimuth difference between the direction of the gap evaluation section and the direction in which the shielding object LN extends is θ, then P3×COS(θ) becomes the evaluation value of the gap in the shielding object LN. However, since the approaching object GG3 intersects with the shielding object LN at a shallow angle, the width of the gap required when it is assumed that the approaching object GG3 passes through the shielding object LN is wider than the vehicle width. If the azimuth difference between the traveling direction X5 of the approaching object GG3 and the direction in which the obstruction LN extends is η, and the width of the gap in the obstruction LN is W, if W×SIN(η) is smaller than the vehicle width, the detected vehicle C6 will not pass through the gap P3. The width of the gap may be evaluated on the assumption that the host vehicle C1 will pass through. In other words, the width of the gap may be evaluated based on the traveling direction X4 of the host vehicle C1. Since the angle of the gap P3 is large with respect to the traveling direction X4 of the host vehicle C1, it can be said that the host vehicle C1 is more likely to pass through the gap P3 than the vehicle C6. The direction based on which the degree of obstruction is evaluated may be selected according to the situation. For example, if the host vehicle is scheduled to park in front of the obstruction LN, there is no need to evaluate the possibility that the host vehicle C1 will pass through the gap, and only the possibility that the approaching object will pass through may be evaluated. In other words, the traveling direction of the vehicle as the reference is selected, and the component perpendicular to the traveling direction is evaluated. The sensor control unit 16 compares the component of the width of the gap evaluation section P3 perpendicular to the reference direction with a gap threshold (e.g., 1.8 m) based on the vehicle width, and evaluates the shielding effect as high if the gap threshold is not met, and evaluates the shielding effect as low if the gap threshold is met or more. Note that when the detection point group FF3 is not detected, the two detection points sandwiching the evaluation position E1 are not aligned, so the width of the gap evaluation section P3 may be applied with, for example, 10 m and the shielding effect may be evaluated as zero, or the shielding effect may be evaluated as zero when the two detection points are not aligned. When there are other gaps, such as a gap sandwiching the evaluation position E2 in the direction in which an approaching object is detected, the shielding effect may be evaluated with the width of the widest gap among them.However, since the approaching object is highly likely to pass through the shielding object LN from its traveling direction, the shielding effect may be evaluated by placing emphasis on the gap in the traveling direction of the approaching object, or only the gap in the traveling direction of the approaching object may be evaluated from the beginning. When evaluating the shielding effect by placing emphasis on the gap in the traveling direction, the shielding effect of multiple gaps may be evaluated individually, and the evaluation value of each shielding effect may be weighted according to the azimuth difference with the traveling direction of the vehicle or the approaching object, and the shielding effect of the entire shielding object LN may be evaluated by the weighted average value of the shielding effect of the multiple gaps. At this time, in order to reduce the amount of calculation, gaps between detection points in the same detection point group may be excluded from the evaluation target of the shielding effect, or may be uniformly given an upper limit value (e.g., 100) for the evaluation value and processed. Gaps between different detection point groups may also be processed in the same manner if they are narrower than a predetermined threshold (e.g., 0.5 m corresponding to the vehicle width of a two-wheeled vehicle).

[0119] Next, a method for determining a detection point group will be described with reference to Fig. 13. Fig. 13 is a diagram illustrating an example of the relationship between detection points and detection point groups. Fig. 13 illustrates the ranges of detection point groups CC4, DD4, and FF4, and detection points (black circles) belonging to each detection point group.

[0120] When a plurality of detection points are detected by the sonar 12, the sensor control unit 16 determines that two or more detection points whose distance between them is equal to or less than a predetermined detection point group threshold belong to the same detection point group, and assigns the same group number (an example of information for identifying a detection point group) to the detection points. When the distance between the detection points exceeds a predetermined detection point group threshold, a unique group number that is not assigned to other detection point groups is assigned. Processing is carried out according to this rule, and when a detection point with a group number and a detection point without a group number are at a distance equal to or less than the detection point group threshold, the latter detection point is assigned the former group number, and when the distance exceeds the detection point group threshold, another unique group number is assigned. When the discrimination process of the detection point group progresses and when the number of detection points increases due to repeated detection, if the distance between two detection points with different group numbers becomes equal to or less than the detection point group threshold, it is determined that the two detection point groups have been integrated, and the detection point groups are merged by overwriting the group number of the detection point belonging to the detection point group with the larger group number with the smaller group number. Here, the detection point cloud threshold may be a value smaller than 1.8 m, which corresponds to the width of a standard four-wheeled vehicle, for example, 0.5 m, which corresponds to the width of a two-wheeled vehicle, or 0.9 m, which corresponds to half the width of a four-wheeled vehicle.

[0121] For example, in the example shown in Fig. 13, the sensor control unit 16 calculates the distances between the four detection points CC41, CC42, CC43, and CC44 detected by the sonar 12. In Fig. 13, the distance between the detection points CC41 and CC42 is L41, the distance between the detection points CC42 and CC43 is L42, and the distance between the detection points CC43 and CC44 is L43. The sensor control unit 16 determines that the four detection points CC41 to CC44 are the same detection point group CC4 based on the fact that the calculated distances L41 to L43 are equal to or less than a predetermined distance E42. In the same manner, the sensor control unit 16 executes group discrimination processing of the detection points included in the two detection point groups DD4 and FF4.

[0122] The sensor control unit 16 also calculates a distance L44 that is the shortest distance between a detection point included in the detection point group CC4 and a detection point included in the detection point group DD4, and determines that the detection point groups CC4 and DD4 are not the same detection point group based on the fact that the calculated distance L44 is greater than the detection point group threshold. Similarly to the distance L44, the sensor control unit 16 also calculates a shortest distance L45 between a detection point included in the detection point group CC4 and a detection point included in the detection point group FF4, and a shortest distance L46 between a detection point included in the detection point group DD4 and a detection point included in the detection point group FF4, and determines that the detection point groups CC4, DD4, and FF4 are different detection point groups based on the fact that each of these distances is greater than the detection point group threshold.

[0123] In the example shown in FIG. 13, the shielding object is estimated on a line passing through the detection point group CC4 and the detection point group DD4. Since the gap between the detection point group CC4 and the detection point group DD4 is wider than the vehicle width (1.8 m) shown in the predetermined distance E42, the shielding effect is evaluated as low (for example, an evaluation value of 60 compared to a maximum value of 100). The detection point group FF4 is between the detection point group CC4 and the detection point group DD4, but the contribution to the shielding effect may be evaluated as zero based on the fact that the shortest distance (L45, L46) between the detection point group CC4 and the detection point group DD4 is greater than the length corresponding to the vehicle length (for example, 4 m). Conversely, if both the shortest distance L45 and the shortest distance L46 are shorter than the vehicle width (1.8 m), the shielding effect may be evaluated as the maximum value (for example, 100). When the shortest distance L45 and the shortest distance L46 are greater than or equal to the vehicle width but less than the vehicle length, the evaluation value of the shielding effect may be set to the larger of the shortest distance L45 and the shortest distance L46, or a value between 60 and 100, for example, depending on the value of the gap closer to the approaching object.

[0124] Next, a method of evaluating the safety level when the parking target position GL1 of the vehicle C1 is set will be described with reference to Fig. 14. Fig. 14 is a diagram for explaining a method of evaluating the safety level when the parking target position GL1 is set.

[0125] (Setting process of target parking position) Here, the parking target position GL1 is the parking target position where the vehicle C1 is parked. The parking target position GL1 is a rectangular area set based on the white lines indicating the parking position drawn in the parking lot. When parking, the driver checks the parking frame drawn in white lines on the road surface when the vehicle C1 is positioned in front of the entrance of the parking target position. At this time, the white lines drawn in the parking lot are simultaneously captured by the camera 11 provided in the right side mirror of the vehicle C1. The camera 11 outputs the captured image to the sensor control unit 16. The sensor control unit 16 performs image processing on the captured image output from the camera 11 to detect a pair of white lines drawn in the parking lot, and determines that the pair of white lines is an indication of a parking frame based on the length and interval of the pair of white lines. When the driver moves the vehicle C1 from in front of the entrance of the parking target position to the position of the vehicle C1 in FIG. 14 and changes the traveling direction from forward to backward, the sensor control unit 16 determines that the parking operation into the previously detected parking space has started, and sets the parking target position GL1 of the vehicle C1 based on the position of the parking space indicated by the pair of white lines.

[0126] (Method of evaluating safety when setting target parking position) In the example shown in FIG. 14, the vehicle C1 moves backward in the traveling direction X6 and attempts to park at the parking target position GL1. When parking between two parked vehicles C7 and C9, the sensor control unit 16 specifies the gap E51 between the detection points in the direction of the parking target position GL1 (the traveling direction X6) as an evaluation target for the shielding effect based on the detection information of the detection points detected by the sonar 12. The gap E51 is a gap evaluated by evaluating the distance from the axis of the traveling direction X6 of the vehicle to the left and right detection points. Since the detection points are lined up along the shield LN, when the gap is evaluated from an oblique direction, it is evaluated to be shorter than when the distance on the shield LN is evaluated. Therefore, the gap E51 evaluated by using the traveling direction X6 of the vehicle as an axis is narrower than the width of the vehicle. In this way, when parking, the vehicle approaches the parking target position from an oblique direction, so if the gap at the parking target position (the gap for one vehicle between the parked vehicles C7 and C9) is evaluated based on the traveling direction of the vehicle, it may be evaluated as an impassable gap. However, since the vehicle plans to park through the gap at the parking target position, the evaluation that the gap in the traveling direction is impassable is not actually correct. Therefore, if the far end of the parking target position (the short side on the far side of the target parking frame) is farther away than the shielding object LN, it is determined that there is a gap that the vehicle can pass through at the position of the shielding object LN that corresponds to the frontage of the parking target position, regardless of the evaluation of the gap at the detection point. Since an approaching object can also pass through this gap, if the approaching object is in a direction based on the parking target position (referring to the range from the frontage of the parking target position GL1 to the far end of the parking target position GL1, that is, GL11), the shielding effect needs to be evaluated low.

[0127] The sensor control unit 16 identifies the position of the parking target position GL1, particularly the position of the short side GL11 (the far end of the parking target position GL1) on the side farther from the host vehicle C1. This position (the position based on the parking target position) is the position where the front end of the vehicle advances. In the collision determination, the determination is made on the condition that the front end of the vehicle does not collide. Therefore, the sensor control unit 16 compares the far end of the parking target position GL1 with the positions of the detection point groups CC5 and DD5 (which may also be the position of the shielding object LN) to evaluate the shielding effect. At the time when the parking target position GL1 is determined, the short side GL11 (the far end of the parking target position GL1), which is the arrival position of the rear end of the host vehicle, is also determined. Therefore, it is only necessary to compare it with the positions of the detection point groups CC5 and DD5.

[0128] When the position of the short side GL11 is farther and separated from the positions of the plurality of detection point groups CC5 and DD5, and rather close to the short side on the front side of the parking target position GL1, regardless of the evaluation of the gap E51, the sensor control unit 16 determines that the host vehicle C1 can pass through the gap between the plurality of detection point groups CC5 and DD5 and evaluates the shielding effect of the shielding object LN as low. That is, if the parking target position is at a position beyond the shielding object LN, there is a gap through which the host vehicle can pass in the shielding object LN. When the host vehicle advances beyond the shielding object LN to the parking target position GL1, the shielding object LN does not shield the host vehicle from approaching objects. Therefore, it is not possible to highly evaluate the shielding effect. In this case, until the host vehicle C1 exceeds the shielding object LN, an evaluation value of the shielding effect in the case where there is a gap exceeding the vehicle width may be given, and the evaluation value of the shielding effect may be set to zero after the host vehicle exceeds the shielding object LN, or the evaluation value may be set to zero from the beginning. On the other hand, when the position of the short side GL11 (the far end of the parking target position GL1) is not farther (is closer) than the positions of the plurality of detection point groups CC5 and DD5 (the line of the shielding object LN), since the plurality of detection point groups CC5 and DD5 on the parking path act as shielding objects until parking at the parking target position GL1, there is no need to lower the evaluation value of the shielding effect. The shielding effect at this time may be evaluated based on the gap between the detection point groups CC5 and DD5 with respect to the traveling direction of the approaching object as in the example described with reference to FIG. 12 above.

[0129] In addition, when the radar 13 detects the approaching object C8, the sensor control unit 16 judges whether or not the traveling direction X6 of the vehicle C1 and the position (direction) of the approaching object C8 relative to the vehicle C1 are substantially the same (specifically, whether or not the direction is a direction in which the vehicle C1 and the approaching object C8 will collide). In the ghost judgment, the ghost likelihood may be judged to be low on the condition that the approaching object is in a direction based on the parking target position GL1. Specifically, the direction based on the parking target position GL1 refers to the short side of the rectangular parking target position GL1 that is closer to the vehicle, that is, the direction of the frontage of the parking frame. This is because there is no reflecting object at the frontage of the parking target position GL1, so a mirror ghost does not occur. Since there is not much difference between the direction of the parking target position and the direction of the frontage of the parking target position, hereinafter, the direction based on the parking target position may be simply called the direction of the parking target position. In the arrangement of FIG. 14, the radar wave may be reflected by the side of the vehicle C7 parked in the adjacent parking space, and may be detected as a multiple reflection by the approaching object C8, but the direction in which the approaching object is detected is still the direction of the parking target position GL1. In other words, if the approaching object is in the direction of the parking target position GL1, the ghost likelihood is determined to be low, so the safety level is not evaluated high or the approaching object is not excluded from the collision determination based on the ghost determination. When the parking target position GL1 is on the other side of the obstruction LN and the traveling direction X6 of the host vehicle C1 and the position (direction) of the approaching object C8 relative to the host vehicle C1 are roughly consistent, the sensor control unit 16 evaluates the safety level of the approaching object C8 as low. This is because the parking target position GL1 is on the other side of the obstruction LN, so it is certain that there is a gap in the obstruction LN through which the vehicle can pass, and when the host vehicle advances to the parking target position GL1, the obstruction LN is no longer an obstruction between the host vehicle and the approaching object, and the obstruction effect becomes zero. Furthermore, the sensor control unit 16 calculates the time margin until collision based on the distance and approaching speed of the approaching object C8 detected by the radar 13, and judges whether emergency braking is necessary or not based on the time margin. In contrast to the above, if the direction in which the approaching object is detected is, for example, the direction of the vehicle C9 that is a parked vehicle on the right of the parking target position GL1, it is expected that the approach will be blocked by the obstruction LN (high obstruction effect), and the estimated ghost position will overlap with the vehicle C9 (low ghost likelihood), so the safety level may be evaluated as high.

[0130] In this way, the sensor control unit 16 can evaluate the shielding effect on the assumption that the vehicle will move to the parking target position GL1 by evaluating the positions of the detection point group detected by the sonar 12 and the positions of obstacles estimated from the detection point group with the parking target position GL1 as a reference. Also, the sensor control unit 16 can appropriately evaluate the degree of safety against an approaching object detected during parking by evaluating the direction in which the approaching object is detected with the direction of the parking target position GL1 as a reference.

[0131] Next, the automatic estimation process of the parking target position GL2 of the vehicle C1 will be described with reference to Fig. 15. Fig. 15 is a diagram for explaining the positional relationship between the parking target position GL2 and the vehicle. In Fig. 15, the vehicle advances in front of the parking target position GL2 in the direction of travel X7 to reach the position of C1, where it stops temporarily, then retreats in the direction of travel X8, and parks between the two parked vehicles C10 and C11.

[0132] Based on various information (e.g., the steering angle, gear position, etc. of the vehicle C1) output from the mechanical sensor 10, the sensor control unit 16 estimates that the vehicle C1 will be parked at the parking target position GL2 when it determines that the direction of travel of the vehicle C1 in the parking lot has changed from forward (direction of travel X7) to reverse (direction of travel X8) and that the steering angle is equal to or greater than a predetermined angle.

[0133] (Automatic estimation process of target parking position) Specifically, when the traveling direction of the vehicle C1 changes from forward (traveling direction X7) to reverse (traveling direction X8), the sensor control unit 16 determines that the vehicle C1 is in a parking lot and is backing up for parking if the vehicle C1 is in an area other than a road (for example, within the premises of a building, store, etc., or the premises of a parking lot operator) based on the position information of the vehicle C1 output from the navigation system 18. The sensor control unit 16 may determine that the vehicle C1 is in a parking lot when it detects a white line corresponding to a parking space from an image captured by the camera 11. The sensor control unit 16 may determine that the vehicle C1 is in a parking lot when it detects that vehicles are lined up at approximately regular intervals using side sonars (sonars FRS, FLS, BRS, BLS) provided on the sides, or may determine that the vehicle C1 is in a parking lot using multiple methods. The above detection may be performed when the vehicle starts to back up, or may use information detected when the vehicle passes in front of a parking space. If the vehicle C1 detects a parking space while moving forward in the parking lot, the information of the detected parking space can be used when the vehicle C1 starts to move backward. When the sensor control unit 16 determines that the vehicle C1 is in the parking lot, the sensor control unit 16 acquires information such as the direction of the vehicle C1's body or the traveling direction of the vehicle C1 until it stops (the traveling direction X8 in the example shown in FIG. 15) based on various information output from the mechanical sensor 10, and estimates the position of the parking target position GL2 based on the information.

[0134] The sensor control unit 16 estimates that the position of the host vehicle C1 when the host vehicle C1 moves forward until the direction of the host vehicle C1 is turned 90° at the current steering angle based on the direction of the host vehicle C1's body when the host vehicle C1 is stopped or the traveling direction of the host vehicle C1 until the host vehicle C1 is stopped is the parking target position GL2. For example, when the steering angle becomes equal to or larger than a predetermined angle, the sensor control unit 16 estimates that the parking operation of the host vehicle C1 will end at a position C1C where the direction of the host vehicle C1's body is turned 90° in the steering angle direction as shown in Fig. 15.

[0135] The sensor control unit 16 calculates the start point STT and end point END of the movement trajectory of the host vehicle C1 when the host vehicle C1 moves in the traveling direction X8 until the direction of the body of the host vehicle C1 turns 90° around the turning center RT1 at the steering angle at the position C1 of the host vehicle, and calculates a rectangular area EXAR with the start point STT as two diagonal points. The detection points and obstacles in the area EXAR have an obstructing effect when the host vehicle is at the position C1, but do not have an obstructing effect when the host vehicle advances to the parking target position GL2. In the example shown in FIG. 15, the host vehicle C1 is parked backward, but the same applies to the case of forward parking. In the example shown in FIG. 15, the start point STT and end point END of the movement trajectory are the positions of the right rear wheels of the host vehicle C1. The method of estimating the position of the parking target position GL2 is not limited to the above, and may be estimated from a line passing through the sonar detection points (CC6, DD61, DD62) and the vehicle length of the host vehicle. When the detection point of the sonar has a gap between CC6 and DD61 and the vehicle starts to move backward toward the gap, it may be estimated that the vehicle moves forward in a direction perpendicular to the line connecting CC6 and DD61 to a position where the vehicle length of the vehicle moves forward. When a white line corresponding to the entrance of the parking frame is detected from the image captured by the camera 11, it may be estimated that the vehicle moves forward in a direction perpendicular to the white line corresponding to the entrance of the parking frame to a position where the vehicle length of the vehicle moves forward. The angle at which the vehicle body of the vehicle C1 turns during parking is not limited to 90°, and the angle of the detected parking frame line relative to the aisle direction or the angle between the short side and the long side of the parking frame line may be applied. The angle may be based on the traveling direction (traveling direction X8) when traveling along the aisle, rather than the orientation of the vehicle body at the start of moving backward. The above is a method for estimating the parking target position when parking manually, but when parking using the automatic parking function, the position of the target parking frame set during the automatic parking process may be used as the parking target position as it is. In addition, the entrance of the parking space may be detected by sonar instead of a camera. In the case of detection by sonar, when the vehicle starts to reverse toward a section without a detection point, it is assumed that the section without a detection point is the entrance of the parking space and that the parking space continues beyond the section.

[0136] (Method of evaluating safety when automatically setting target parking position) When the parking target position GL2 is estimated, the sensor control unit 16 may perform the evaluation of the safety level by excluding the detection points (detection points DD61 and DD62 in the area EXAR in the example shown in FIG. 15) located between the parking target position GL2 and the far end of the parking target position GL2, on the assumption that the vehicle C1 advances to the parking target position GL2. The exclusion of the detection point CC6 may be determined according to the direction of the approaching object. For example, when the vehicle is in C1, if the approaching object comes from a direction of 90 degrees to the right (the direction of the vehicle C11), the detection point CC6 may be excluded at the same time as DD61, and if the approaching object is from the rear right (the direction of the vehicle C10), the detection point CC6 is considered to have a shielding effect even after the vehicle advances to the parking target position GL2, so it should not be excluded from the evaluation of the safety level.

[0137] Next, a method for evaluating the ghost likelihood and the safety level when the parking target position GL3 of the vehicle C1 is estimated will be described with reference to Fig. 16 and Fig. 17. Fig. 16 and Fig. 17 are diagrams showing the vehicle, the parking target position GL3, the positions of the detection points, and the estimated positions of the obstruction. The vehicle C1 shown in Fig. 16 and Fig. 17 moves backward in the traveling direction X9 and parks at the parking target position GL3.

[0138] (Method of evaluating ghost likelihood and safety when setting a parking target position) The following description will be given with reference to Fig. 16. The sensor control unit 16 estimates an obstruction on a line LNA that connects multiple detection points detected by the sonar 12. The sensor control unit 16 assumes that a ghost is generated by the estimated obstruction, and estimates the estimated ghost position at a position symmetrical to the vehicle C1, with the obstruction position (line LNA) as the axis of symmetry. Also, it is assumed that the radar detects an approaching object C13 at the same position as the estimated ghost position. In other words, since the approaching object C13 is detected at the same position as the estimated ghost position, it may be a ghost or an actual approaching vehicle.

[0139] The sensor control unit 16 calculates the position of the short side GL31 of the parking target position GL3 and the position of the approaching object C13. The sensor control unit 16 compares the distance from the host vehicle C1 to the position of the short side GL31 and the position of the approaching object C13, and determines whether the approaching object C13 is farther away than the position of the short side GL31.

[0140] When the sensor control unit 16 determines that the approaching object C13 is closer than the position of the short side GL31 or is at the same distance, it evaluates the ghost likelihood for the approaching object as high. However, the rule described with reference to FIG. 14 that determines the ghost likelihood as low on the condition that the approaching object is in the direction of the parking target position takes precedence. In the case of FIG. 16, the approaching object is not in the direction of the parking target position, so it is not subject to the above rule. A position closer to the short side GL31, which is the far end of the parking target position, is less likely to be able to accommodate a vehicle, so it may be evaluated as having a high ghost likelihood.

[0141] In the case of FIG. 16, there is a parked vehicle to the right of the parking target position GL3, and the detection points CC7 and DD7 are detections of the parked vehicle, and it is possible that the radar wave is reflected by the parked vehicle, resulting in the ghost of the approaching object C13 (the approaching object C13 is a ghost). If there is a parked vehicle corresponding to the detection points CC7 and DD7, the approaching object C13 overlaps with the parked vehicle, so it can be said that the approaching object C13 is unlikely to have a real body (the ghost likelihood is high). Alternatively, it can be considered that there is a guardrail to the right of the parking target position GL3, and the radar wave is reflected by the guardrail, resulting in the ghost at the position of the approaching object C13. In this case, too, there is a linear structure such as a guardrail corresponding to the detection points CC7 and DD7, and it is unlikely that a vehicle actually exists at the position of the approaching object C13 behind it (that is, it is highly likely that the approaching object C13 is a ghost). This is because, although the vehicle can be parked at the parking target position GL3 and there is a space to the right of the parking target position GL3, there is no rational reason for providing a guardrail or the like that would prevent the vehicle from passing at the detection points CC7 and DD7. Therefore, the sensor control unit 16 may estimate the ghost likelihood of the approaching object to be a ghost higher when it is determined that the approaching object is closer than the position of the short side GL31 of the parking target position GL3 than the ghost likelihood when it is determined that the position of the approaching object is farther than the position of the short side GL31. For example, when the approaching object is located farther than the line LNB, there is a possibility that the vehicle is traveling on a passage provided on the other side of the line of vehicles to be parked, so it is appropriate to evaluate the ghost likelihood lower than when the approaching object is located between the lines LNA and LNB. Note that when the approaching object is in the direction of the parking target position, the approaching object is located at a position where the vehicle may actually be present, so the ghost likelihood is evaluated lower even if it is between the lines LNA and LNB. In other words, if the approaching object is closer than the far end of the parking target position, the rule that evaluates the ghost likelihood lower when the approaching object is in the direction of the parking target position takes precedence over the rule that increases the ghost likelihood compared to when the object is farther away. In the case of Figure 16, since the estimated ghost position is at C13, it does not correspond to the case where the approaching object is in the direction of the parking target position, and the approaching object is closer than the position of the short side GL31, the ghost likelihood is evaluated low.

[0142] FIG. 17 is a diagram for explaining the relationship between the parking target position GL3, the position of the detection point group, the shielding effect, and the ghost likelihood. It is assumed that an approaching object (not shown) is located farther away from the line LNA as viewed from the vehicle C1. In the example shown in FIG. 17, the detection points CC7 and FF7 detected by the sonar 12 are located closer to the vehicle C1 than the approaching object, and therefore may function as a shield that prevents the approaching object from advancing to the position of the vehicle C1. However, when the parking target position GL3 is estimated, the sensor control unit 16 estimates that the shielding object on the line LNA indicated by the detection points CC7 and FF7 is a shielding object having a gap for the vehicle C1 to park at the parking target position GL3, and determines it by discounting the shielding effect. At the same time, since the shielding object on the line LNA is highly likely to be another vehicle or a wall, if the approaching object is located at the position of C13, the ghost likelihood is evaluated as high, but if the approaching object is located at the parking target position GL3, the ghost likelihood is estimated as low. Therefore, if the evaluation value of the shielding effect is low and the ghost likelihood is also low, the evaluation value of the safety level will be low. In other words, if there is an approaching object near the parking target position GL3, the automatic brake is more likely to be activated.

[0143] Conversely, in Fig. 17, if there are no detection points CC7 and DD7, but there are detection points detected by the sonar 12 at the positions of detection points GG71 and GG72 (i.e., on the line LNB shown in Fig. 17), and the approaching object is located farther than the line LNB, the sensor control unit 16 presumes that the detection points GG71 and GG72 are detection information of an obstacle present behind the parking target position GL3 (i.e., on the other side of the parking target position GL3 as viewed from the host vehicle C1). In such a case, the sensor control unit 16 may evaluate the safety level as high because the obstacle is located farther than the short side GL31 of the parking target position GL3 and has a shielding effect that blocks approaching objects from a greater distance.

[0144] In addition, if there is a group of detection points arranged in a line behind the parking target position GL3, the probability that the approaching object is a structure such as a wall is high, so if the approaching object is located farther away than the group of detection points arranged in a line, the ghost likelihood may be evaluated higher than if the approaching object is located in front of the parking target position GL3.

[0145] Next, a method for evaluating the ghost likelihood and the safety level when the parking target position GL3 of the vehicle C1 is estimated will be described with reference to Fig. 18. Fig. 18 is a diagram for explaining the parking target position GL4 and the positions of the detection point group. The vehicle C1 shown in Fig. 18 retreats in the traveling direction X10 and parks at the parking target position GL4.

[0146] The sensor control unit 16 executes detection point group determination for the multiple detection points detected by the sonar 12. In the example shown in Fig. 18, the multiple detection points detected by the sonar 12 are classified into three detection point groups CC8, DD8, and EE8 based on the distance between the detection points.

[0147] In the evaluation of the safety level, the sensor control unit 16 determines that the detection point group CC8 located in the direction of travel X10 from the current position toward the parking target position GL4 has a large contribution to the safety of the vehicle C1, and the detection point group DD8 not located in the direction of travel X10 has a relatively small contribution to the safety of the vehicle C1. In such a case, the sensor control unit 16 increases the weighting of each detection point included in the detection point group CC8 and decreases the weighting of each detection point included in the detection point group DD8 in the evaluation process of the safety level. Note that, in the evaluation process of the safety level, the sensor control unit 16 may execute the evaluation process of the safety level using only the detection points included in the detection point CC7 that are close to the parking target position GL4. In addition, since the safety level described here is obtained by the shielding effect of an obstacle, the above-mentioned safety level may be replaced with the shielding effect. In addition, the evaluation of the safety level may be based not only on the traveling direction of the vehicle but also on the direction in which an approaching object is detected. For example, when the radar of the vehicle C1 detects an approaching object on the left rear side, the detection point group EE8 may be added to the evaluation of the safety level. The detection point group EE8 detects the presence of the adjacent vehicle C14, and once the vehicle is parked at the parking target position GL4, it can be determined that it is safe to continue parking, since the adjacent vehicle C14 will act as a shield against objects approaching from the left.

[0148] Further, the sensor control unit 16 estimates that the detection point group CC8 and the detection point group DD8 indicate an obstacle on the line LN2 located behind the parking target position GL4 (that is, on the other side of the parking target position GL4 as seen from the vehicle C1). In such a case, the sensor control unit 16 may calculate the estimated ghost position assuming that the line LN2 is a reflective surface and evaluate the ghost likelihood for an approaching object farther than the line LN2, but if the approaching object is closer than the line LN2, it cannot be a ghost, so the ghost likelihood evaluation may be omitted. The safety level may be evaluated high without evaluating the ghost likelihood because the vehicle C1 does not move to a position farther than the short side GL41 of the parking target position GL4 and is protected by the detection point (obstacle). Also, in the collision judgment, if the vehicle C1 intersects with the path of the approaching object farther than the line LN2, it may be judged that there will be no collision on the premise that the vehicle C1 stops at the parking target position GL4.

[0149] The relationship between the detection of an object by the sonar 12 and the detection of an approaching object by the radar 13 and the evaluation of the ghost likelihood and the evaluation of the safety level will be described with reference to Fig. 19. Fig. 19 is a diagram showing the arrangement of the detected objects and the approaching objects. Fig. 19 shows an example in which the host vehicle C1 is backed up in the traveling direction X12 and parked at the parking target position GL5.

[0150] The ghost likelihood is used to evaluate the safety level. When the ghost likelihood is highly evaluated, since the approaching object is likely to be a ghost and has a low possibility of colliding with the host vehicle, the safety level is highly evaluated. Also, when the ghost likelihood is lowly evaluated, since the approaching object is likely not to be a ghost and may collide with the host vehicle, the safety level is lowly evaluated. Further, when the sensor control unit 16 determines that there is an obstacle that can shield the approach of the approaching object, it is presumed that the approaching object has a low possibility of approaching and colliding with the host vehicle due to this obstacle, and the safety level is highly evaluated. Also, when the sensor control unit 16 determines that there is no obstacle that can shield the approach of the approaching object, it is presumed that the approaching object may approach and collide with the host vehicle due to this obstacle, and the safety level is lowly evaluated. That is, the safety level is highly evaluated when the sensor control unit 16 determines that the ghost likelihood for the approaching object detected by the radar 13 is high or there is an obstacle that can shield the approach of the approaching object.

[0151] When the safety level is high, the approaching object detected by the radar 13 is excluded from the determination of the necessity for emergency braking (i.e., collision determination). On the other hand, the detected object (obstacle) detected by the sonar 12 is the subject of the determination of the necessity for emergency braking (collision determination) regardless of the ghost likelihood.

[0152] For example, in the example shown in FIG. 19, when the radar 13 of the host vehicle C1 detects an approaching object C17, the sensor control unit 16 may evaluate that there is an obstacle (detection point) OB2 detected by the sonar 12 and located on the traveling direction X12 of the host vehicle C1, and thus there is an obstacle that can shield the approach of the approaching object C17 to the host vehicle C1. Based on this evaluation of the shielding degree, the sensor control unit 16 may highly evaluate the safety level for the approaching object C17 detected by the radar 13.

[0153] However, the object evaluated by the sensor control unit 16 as having a high safety level is only the approaching object C17 detected by the radar 13, and not the detected object (obstacle OB2) detected by the sonar 12. Therefore, even if the sensor control unit 16 evaluates the approaching object C17 as having a high safety level, it does not exclude the obstacle OB2 detected by the sonar 12 from the safety level evaluation targets (i.e., targets for collision determination), and performs the safety level evaluation process for the obstacle OB2 as well.

[0154] Since the obstacle OB2 is located in front of the parking target position GL5 (i.e., in the direction in which the host vehicle C1 is approaching), the sensor control unit 16 executes a collision determination as to whether or not the obstacle OB2 will collide with the host vehicle C1. When the sensor control unit 16 evaluates in the collision determination that there is a risk (possibility) of the obstacle OB2 colliding with the host vehicle C1 if the vehicle continues to move backward, the sensor control unit 16 causes the vehicle control unit 17 to execute emergency braking. At this time, the collision risk with the approaching object C17 may be added to the calculation, taking into account the possibility that the obstacle OB2 is not a fixed object, but is an object such as a cone placed on the ground that cannot prevent the approach of the approaching object C17. This is an addition that adds to the evaluation value of the risk, so even if it is added, the conclusion to execute emergency braking does not change. Alternatively, since the vehicle is trying to park on the other side of the obstacle OB2, it may be presumed that the obstacle OB2 does not exist or is an object (such as a step) that does not block the approaching object C17, and the degree of blocking by the detection point on LN3 may be evaluated low, and the safety degree of the approaching object C17 may be evaluated low. In other words, even in this case, if the processing is appropriate, the conclusion that emergency braking should be performed remains the same.

[0155] An example of an operation procedure of the host vehicle C1 according to the first embodiment will now be described with reference to Fig. 20. Fig. 20 is a flowchart showing an example of an operation procedure of the host vehicle C1 according to the first embodiment.

[0156] First, the sensor control unit 16 in the host vehicle C1 causes the twelve sonars 12 to detect objects (obstacles, etc.) within their respective detection ranges, and causes the three radars 13 to detect reflectors (approaching objects, etc.) within their respective scanning ranges.

[0157] The twelve sonars output detection information on detected objects to the sensor control unit 16. In addition, the three radars 13 output detection information on detected reflectors to the sensor control unit 16. The above has already been performed before START, and is not shown in the figures.

[0158] The sensor control unit 16 determines whether or not there is an object detected by the sonar 12 and the distance between the position of the object and the vehicle C1 is equal to or less than a predetermined distance threshold (for example, 3 m) (St11).

[0159] When the sensor control unit 16 determines in the process of step St11 that the distance between the position of the detected object and the vehicle C1 is equal to or shorter than a predetermined distance (St11, YES), the process proceeds to collision determination process from St11 onward. The collision determination process from St12 onward will be described later.

[0160] On the other hand, if the sensor control unit 16 determines in the processing of step St11 that there is no detected object by the sonar 12 or that the distance between the position of the detected object and the vehicle C1 exceeds a predetermined distance threshold (St11, NO), the sensor control unit 16 determines whether or not an approaching object within a predetermined distance has been detected by the radar 13. If an approaching object is detected by the radar 13, the sensor control unit 16 determines whether or not the distance between the position of the approaching object and the vehicle C1 is equal to or less than a predetermined distance (for example, 5 m) (St13).

[0161] If the sensor control unit 16 determines in the processing of step St13 that no approaching object is detected or that the distance between the position of the approaching object and the vehicle C1 is not less than a predetermined distance (St13, NO), it terminates the operation procedure shown in Figure 20.

[0162] On the other hand, when the sensor control unit 16 detects an approaching object and determines that the distance between the vehicle C1 and the approaching object is equal to or less than a predetermined distance in the process of step St13 (St13, YES), it executes evaluation of the ghost likelihood (St14). The process of step St14 is not essential, and the process may proceed to step St15 without evaluating the ghost likelihood. For example, when the sonar detection points are arranged at intervals of equal to or less than a predetermined distance (for example, 1.8 m), indicating the presence of an obstacle that the vehicle cannot pass through, and the obstacle is blocking the direction of the approaching object or the traveling direction of the vehicle C1, it is sufficiently safe even if the approaching object is not a ghost, so that the evaluation of the ghost likelihood may be omitted, or a control procedure may be implemented that does not evaluate the ghost likelihood at all from the beginning. In step St15, the safety level is evaluated from the ghost likelihood (if any) and the degree of blocking by the detection points.

[0163] The sensor control unit 16 judges whether the safety level evaluated in the process of step St15 is less than a predetermined safety level threshold (St16). The predetermined safety level threshold is a value for judging whether emergency braking is necessary or not, and is set to a value corresponding to the maximum safety level. For example, when the maximum safety level is "100", the predetermined safety level threshold is set to a value such as "80" or "75".

[0164] When it is determined in the process of step St16 that the safety level is equal to or higher than the predetermined value (St16, NO), the sensor control unit 16 ends the operation procedure shown in FIG.

[0165] On the other hand, if the sensor control unit 16 determines in the process of step St16 that the safety level is less than the predetermined value (St16, YES), the process proceeds to collision determination process (St12). The same applies if the sensor control unit 16 determines in step St11 that the object detected by the sonar 12 is in close range (St11, YES). In step St12, the sensor control unit 16 estimates the movement trajectory of the host vehicle C1 from the current time until a predetermined time (e.g., 6 seconds) later, and the movement trajectory of the approaching object from the current time until a predetermined time (e.g., 6 seconds) later.

[0166] In the next step St17, the sensor control unit 16 calculates a first time margin until a collision occurs between the host vehicle C1 and the detected object or the approaching object based on the estimated movement trajectory of the host vehicle C1 and the movement trajectory of the detected object or the approaching object from the current time until a predetermined time (e.g., 6 seconds later).

[0167] The sensor control unit 16 determines whether the calculated first time margin is less than a first time threshold (for example, 5 seconds) (St18).

[0168] If the calculated first time margin is equal to or greater than the first time threshold in the process of step St18 (St18, NO), the sensor control unit 16 reserves the emergency braking and ends the operation procedure shown in FIG.

[0169] On the other hand, if the sensor control unit 16 determines in the processing of step St18 that the calculated first time margin is less than the first time threshold (St18, YES), it calculates a second time margin until a collision between the host vehicle C1 and a detected object or an approaching object when deceleration control is performed by the vehicle control unit 17 (i.e., the traveling speed of the host vehicle C1 is decelerated) (St19).

[0170] Next, the sensor control unit 16 determines whether or not the second time margin calculated in the process of step St19 is less than a second time threshold value (for example, 4 seconds) (St20).

[0171] When the sensor control unit 16 determines that the time margin calculated in the process of step St19 is less than a second time threshold value (e.g., 4 seconds) (St20, YES), it generates a control command requesting emergency braking and outputs it to the vehicle control unit 17. The vehicle control unit 17 executes emergency braking based on the control command output from the sensor control unit 16 (St21).

[0172] On the other hand, when the sensor control unit 16 determines that the second time margin calculated in the process of step St19 is equal to or greater than a second time threshold (for example, 4 seconds, etc.) (St20, NO), it generates a control command requesting deceleration control and outputs it to the vehicle control unit 17. The vehicle control unit 17 executes deceleration control based on the control command output from the sensor control unit 16 (St22).

[0173] As described above, the vehicle control device 20 mounted on the host vehicles C1, C1A, and C1B (an example of a vehicle) according to the first embodiment includes a sonar 12 or a radar 13 (an example of an acquisition unit) that acquires detection information of an obstacle around the host vehicle, and a sensor control unit 16 that performs a collision determination for evaluating the possibility of collision with the obstacle. Based on the detection information, the sensor control unit 16 generates information on an approaching object that is an obstacle approaching the host vehicle and information on a detection point indicating an immovable obstacle, estimates the position of the shielding object based on the information on the detection point, evaluates a ghost likelihood indicating the possibility that the approaching object is a ghost based on the position of the shielding object and the information on the approaching object, and excludes the approaching object from the collision determination based on the ghost likelihood.

[0174] Thus, the vehicle control device 20 mounted on the host vehicles C1, C1A, and C1B according to the first embodiment is a vehicle control device mounted on the host vehicle, and includes a sonar 12 or a radar 13 that acquires detection information of an obstacle around the host vehicle, and a sensor control unit 16 that performs a collision determination for evaluating the possibility of collision with the obstacle. The sensor control unit 16 generates information on an approaching object that is an obstacle approaching the host vehicle and information on a detection point indicating an immovable obstacle based on the detection information. The sensor control unit 16 estimates the position of the shielding object based on the information on the detection point, evaluates a ghost likelihood indicating the possibility that the approaching object is a ghost based on the position of the shielding object and the information on the approaching object, and excludes the approaching object from the collision determination based on the ghost likelihood. As a result, an approaching object determined to be a ghost with a high ghost likelihood and no danger is excluded from the target of the collision determination, so that unnecessary emergency braking can be avoided.

[0175] Furthermore, as described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment identifies an estimated ghost position where a ghost may occur based on the position of an obstruction, and evaluates the ghost likelihood as high when the larger of the distance between the estimated ghost position and the position of the approaching object, or the orientation discrepancy evaluating the orientation difference between the orientation of the estimated ghost position and the orientation of the approaching object, and the distance discrepancy evaluating the distance difference between the distance of the estimated ghost position and the distance of the approaching object, or the sum of the orientation discrepancy and the distance discrepancy, or the weighted average of the orientation discrepancy and the distance discrepancy is small. As a result, the sensor control unit 16 in the vehicle control device 20 of embodiment 1 identifies an estimated ghost position where a ghost may occur based on the position of an obstruction, and evaluates the ghost likelihood as high when the larger of the distance between the estimated ghost position and the position of the approaching object, or the orientation discrepancy evaluating the orientation difference between the orientation of the ghost position and the orientation of the approaching object, and the distance discrepancy evaluating the distance difference between the distance of the ghost position and the distance of the approaching object, or the sum of the orientation discrepancy and the distance discrepancy, or the weighted average of the orientation discrepancy and the distance discrepancy is small, thereby enabling the ghost likelihood to be evaluated with high accuracy.

[0176] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment generates information on an approaching object from detection information detected by the radar 13. The degree of difference in orientation, the degree of difference in distance, or the weight of the weighted average is based on the standard error in the orientation direction of the radar 13 and the allowable error in the distance direction of the radar 13. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment generates information on an approaching object from detection information by the radar 13, and the degree of difference in orientation, the degree of difference in distance, or the weight of the weighted average is based on the standard error in the orientation direction of the radar 13 and the allowable error in the distance direction of the radar 13, thereby making it possible to accurately evaluate the ghost-likeness according to the error performance of the radar 13.

[0177] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment repeatedly evaluates the ghost likelihood in a time series manner, and excludes the approaching object from the collision judgment when the sum, average, or weighted average of the ghost likelihood evaluated in a time series manner is equal to or greater than a predetermined value. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment repeatedly evaluates the ghost likelihood in a time series manner, and excludes the approaching object from the collision judgment when the sum, average, or weighted average of the ghost likelihood evaluated in a time series manner is equal to or greater than a predetermined value, thereby making it possible to accurately evaluate the ghost-likeliness by using the cumulatively detected detection information.

[0178] As described above, the weight of the weighted average calculated by the vehicle control device 20 according to the first embodiment is based on the number of detection points related to the evaluation of the ghost likelihood or the order of the chronological evaluation. The ghost likelihood evaluated later is evaluated more heavily than the ghost likelihood evaluated earlier. As a result, the vehicle control device 20 according to the first embodiment evaluates the weight of the weighted average based on the number of detection points related to the evaluation of the ghost likelihood or the order of the chronological evaluation, and evaluates the ghost likelihood evaluated later more heavily than the ghost likelihood evaluated earlier, so that the evaluation when the accumulated amount of detection information increases is evaluated more heavily than the evaluation when the accumulated amount of detection information is small, thereby making it possible to accurately evaluate the ghost-likeliness.

[0179] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood lower when there is no detection point in the direction of the approaching object than when there is a detection point in the direction of the approaching object. As a result, the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood lower when there is no detection point in the direction of the approaching object than when there is a detection point in the direction of the approaching object, so that when there is no reflecting object in the direction of the approaching object, it does not determine that it is a ghost, and it is possible to avoid excluding approaching objects that are not ghosts from the collision determination.

[0180] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood lower when the number of detection points in the direction of the approaching object is small than when the number of detection points in the direction of the approaching object is large. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood lower when the number of detection points in the direction of the approaching object is small than when the number of detection points in the direction is large, so that when it is suspected that there is a reflecting object in the direction of the approaching object, it does not determine that it is a ghost, and it is possible to avoid excluding approaching objects that are not ghosts from the collision determination.

[0181] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment generates an approximate straight line LS based on a plurality of detection points, calculates the average distance from the generated approximate straight line LS to the plurality of detection points, or the variance of the plurality of detection points relative to the approximate straight line LS, and evaluates the ghost likelihood highly when the calculated average distance or variance is small. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment generates an approximate straight line based on a plurality of detection points, calculates the average distance from the generated approximate straight line to the plurality of detection points, or the variance of the plurality of detection points relative to the approximate straight line, and evaluates the ghost likelihood highly when the calculated average distance or variance is small, thereby making it possible to avoid excluding an approaching object that is not a ghost from the collision judgment when it is suspected that there is a reflective object in the direction of the approaching object.

[0182] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood as zero when the calculated average distance or variance is equal to or greater than a predetermined threshold. As a result, by evaluating the ghost likelihood as zero when the calculated average distance or variance is equal to or greater than a predetermined threshold, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment does not determine that an approaching object is a ghost when there is no possibility of a reflecting object in the direction of the approaching object, and can avoid excluding approaching objects that are not ghosts from the collision determination.

[0183] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood higher when a large number of detection points are used to generate the approximate straight line LS than when a small number of detection points are used. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood higher when a large number of detection points are used to generate the approximate straight line LS than when a small number of detection points are used, thereby evaluating the ghost likelihood higher when the probability that a reflective surface exists is high, and it is possible to accurately evaluate the ghost-likeliness.

[0184] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood based on the interval between the detection points, and evaluates the ghost likelihood higher when the interval between the detection points is narrow than when the interval between the detection points is wide. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood based on the interval between the detection points, and evaluates the ghost likelihood higher when the interval between the detection points is narrow than when the interval between the detection points is wide, so that when there is a high probability that a reflective surface exists, the ghost likelihood is highly evaluated, and the ghost-likeliness can be evaluated with high accuracy.

[0185] As described above, when there is a parking target position where the vehicle is to be parked, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood to be low if there is an approaching object in a direction based on the parking target position of the vehicle. As a result, when there is a parking target position where the vehicle is to be parked, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood to be low if there is an approaching object in a direction based on the parking target position of the vehicle, thereby evaluating the ghost likelihood on the premise that no reflective surface exists in the direction of the parking position, and can accurately evaluate the ghost-likeliness.

[0186] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood lower when there is a parking target position where the vehicle is parked and the approaching object is closer to the position based on the parking target position of the vehicle than when the approaching object is farther away from the position based on the parking target position. As a result, the vehicle control device 20 according to the first embodiment evaluates the ghost likelihood lower when there is a parking target position where the vehicle is parked and the approaching object is closer to the position based on the parking target position of the vehicle than when the approaching object is farther away from the position based on the parking target position, thereby reflecting the condition that an approaching object having a substance cannot exist in relation to the parking position in the ghost likelihood, thereby making it possible to accurately evaluate the ghost-likeliness.

[0187] As described above, the vehicle control device 20 mounted on the host vehicle C1, C1A, C1B according to the first embodiment includes a sonar 12 or a radar 13 (an example of an acquisition unit) that acquires detection information of an obstacle around the host vehicle (an example of a vehicle), and a sensor control unit 16 that performs collision judgment to evaluate the possibility of collision with the obstacle. Based on the detection information, the sensor control unit 16 generates information on an approaching object that is an obstacle approaching the host vehicle, and information on a detection point cloud that is a collection of detection points that indicate a stationary obstacle, and excludes the approaching object from collision judgment when the detection point cloud has a shielding effect of shielding the host vehicle from the approaching object. The shielding effect is evaluated using a gap threshold based on the vehicle width.

[0188] As a result, the vehicle control device 20 of embodiment 1 can effectively evaluate the shielding effect by which the vehicle is protected by an obstacle by evaluating the shielding effect using a gap threshold based on the vehicle width, and can avoid performing unnecessary emergency braking by excluding approaching objects that are not dangerous from the targets of collision judgment.

[0189] As described above, in the vehicle control device 20 according to the first embodiment, the gap threshold based on the vehicle width is set to a value equal to or smaller than the vehicle width of a passenger vehicle. The sensor control unit 16 compares the gap at the detection point with the gap threshold, and evaluates the shielding effect lower when the gap at the detection point is wider than the gap threshold than when the gap at the detection point is narrower than the gap threshold. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment sets the gap threshold based on the vehicle width to a value equal to or smaller than the vehicle width of a passenger vehicle, compares the gap at the detection point with the gap threshold, and evaluates the shielding effect lower when the gap at the detection point is wider than the gap threshold than when the gap at the detection point is narrower than the gap threshold, thereby making it possible to perform emergency braking when a passenger vehicle can pass through the gap.

[0190] As described above, the vehicle control device 20 according to the first embodiment uses the direction of the approaching object, the traveling direction of the host vehicle, or the traveling direction of the approaching object as the reference direction. The shielding effect is evaluated for the gap at the detection point based on the reference direction. As a result, the vehicle control device 20 according to the first embodiment evaluates the gap at the detection point based on the reference direction, thereby evaluating the distance between the obstacles in the direction in which the host vehicle moves, and can accurately evaluate the possibility that the vehicle will pass through the gap.

[0191] As described above, the vehicle control device 20 according to the first embodiment evaluates gaps between the detection points of the detection point group located in the reference direction, and in evaluating the shielding effect, evaluates gaps with small deviation angles with respect to the reference direction more heavily than gaps with large deviation angles with respect to the reference direction. As a result, the vehicle control device 20 according to the first embodiment evaluates gaps with small deviation angles with respect to the reference direction more heavily than gaps with large deviation angles with respect to the reference direction, thereby placing emphasis on gaps in the direction in which the vehicle moves and not on gaps in the direction in which the vehicle moves, and can accurately evaluate the possibility that the host vehicle will pass through the gap.

[0192] As described above, the vehicle control device 20 according to the first embodiment evaluates the number of detection points of the detection point group located in the reference direction, and evaluates the shielding effect higher when the number of detection points is large than when the number of detection points is small. As a result, the vehicle control device 20 according to the first embodiment evaluates gaps between the detection points of the detection point group located in the reference direction, and in the evaluation of the shielding effect, evaluates gaps with a small deviation angle with respect to the reference direction more heavily than gaps with a large deviation angle with respect to the reference direction, thereby placing importance on gaps in the direction in which the host vehicle moves and not on gaps in the direction in which the host vehicle moves, thereby making it possible to accurately evaluate the possibility that the vehicle will pass through the gap.

[0193] As described above, when there is a parking target position where the host vehicle is to be parked, the vehicle control device 20 according to the first embodiment lowers the evaluation of the shielding effect when an approaching object is in a direction based on the parking target position. As a result, when there is a parking target position where the host vehicle is to be parked and the approaching object is in a direction based on the parking target position, the vehicle control device 20 according to the first embodiment lowers the evaluation of the shielding effect, thereby being able to accurately evaluate the possibility that the host vehicle is shielded by an obstacle on the premise that there is a gap in the direction of the parking target position through which the vehicle can pass.

[0194] As described above, when there is a parking target position where the host vehicle is to be parked, the vehicle control device 20 according to the first embodiment evaluates the shielding effect lower when the detection point cloud is closer than a position based on the parking target position than when it is farther from the position based on the parking target position. As a result, when there is a parking target position where the host vehicle is to be parked, the vehicle control device 20 according to the first embodiment evaluates the shielding effect lower when the detection point cloud is closer than a position based on the parking target position than when it is farther from the position based on the parking target position, thereby evaluating the shielding effect on the premise that the host vehicle will advance to the parking target position, and can effectively evaluate the detection point cloud that has a low shielding effect when the host vehicle advances to the parking target position.

[0195] As described above, the detection information detected by the vehicle control device 20 according to the first embodiment includes the detection information detected by the radar 13. The information of the approaching object is based on the detection information detected by the radar 13. As a result, the vehicle control device 20 according to the first embodiment can utilize the characteristics of the radar 13 that can detect the approaching speed, since the detection information includes the detection information detected by the radar and the information of the approaching object is based on the detection information detected by the radar 13.

[0196] As described above, the evaluation target of the shielding effect of the vehicle control device 20 according to the first embodiment includes the obstacle detected by the radar 13. As a result, the evaluation target of the shielding effect of the vehicle control device 20 according to the first embodiment includes the obstacle detected by the radar 13, and thus the vehicle control device 20 according to the first embodiment can add the detection point in the direction in which the approaching object is detected to the evaluation of the shielding effect, thereby enabling the shielding effect to be accurately evaluated.

[0197] Furthermore, as described above, the detection information detected by the vehicle control device 20 according to the first embodiment includes the detection information detected by the sonar 12. The sensor control unit 16 does not exclude the detection points detected by the sonar 12 from the collision determination. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment includes the detection information detected by the sonar 12, and by not excluding the detection points detected by the sonar 12 from the collision determination, it is possible to ensure that emergency braking is performed for any object detected by the sonar 12.

[0198] As described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment selects detection points to be evaluated for the shielding effect from the detection points detected by the sonar 12 based on the reception strength of the sonar 12. As a result, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment selects detection points to be evaluated for the shielding effect from the detection points detected by the sonar 12 based on the reception strength of the sonar 12, thereby excluding obstacles that cannot shield an approaching object, such as a step, from the evaluation targets for the shielding effect, and prevents emergency braking from being applied only when there is actually a shielding effect.

[0199] Also, as described above, the sensor control unit 16 in the vehicle control device 20 according to the first embodiment generates an approximate straight line LS based on the positions of a plurality of detection points included in the detection point group, calculates the average distance from the generated approximate straight line LS to the positions of the plurality of detection points, or the variance of the positions of the plurality of detection points with respect to the approximate straight line LS, and when the calculated average distance or variance is small, highly evaluates the shielding effect. Thereby, the vehicle control device 20 according to the first embodiment generates an approximate straight line based on the positions of a plurality of detection points included in the detection point group, calculates the average distance from the generated approximate straight line to the positions of the plurality of detection points, or the variance of the positions of the plurality of detection points with respect to the approximate straight line, and when the calculated average distance or variance is small, highly evaluates the shielding effect, so that when there is an obstacle with a high degree of certainty of existence, it is possible to ensure that emergency braking does not work.

[0200] As described above, various embodiments have been described with reference to the drawings. Needless to say, the present disclosure is not limited to such examples. It is obvious that those skilled in the art can conceive of various modification examples, correction examples, substitution examples, addition examples, deletion examples, and equivalent examples within the scope described in the claims, and it is naturally understood that they also belong to the technical scope of the present disclosure. Further, within the scope not departing from the gist of the invention, the components in the above-described various embodiments may be arbitrarily combined.

Industrial Applicability

[0201] The present disclosure is useful as a vehicle control device, a vehicle, a vehicle control method, and a vehicle control program that can accurately determine whether a detected approaching object should be excluded from collision determination.

Explanation of Signs

[0202] 10 Mechanical Sensor 11 Camera 12 Sonar 13 Radar 14 Memory 15 HMI 16 Sensor Control Unit 17 Vehicle Control Unit 18 Navi 19 In-car LAN BLCX,BLSX1,BLSX2,BLX,BRCX,BRCX11,BRCX12,BRSX1,BRSX2,BRSX11,BRSX12,BRX,FLCX,FLSX1,FLSX2,FLX,FRCX,FRSX1,FRSX2,FRX Detection range C1, C1A, C1B Vehicle GL1,GL2,GL3,GL4,GL5 Parking target position GL11,GL31,GL41 Short side L1AR, L2AR, L3AR Scanning Range LS approximate straight line

Claims

1. A vehicle control device mounted on a vehicle, an acquisition unit that acquires detection information of an obstacle detected around the vehicle; a sensor control unit that performs a collision determination to evaluate a possibility of a collision with the obstacle, The sensor control unit, based on the detection information, Information on an approaching object that is an obstacle approaching the vehicle; and Generate information on a detection point cloud, which is a set of detection points indicating stationary obstacles; If the detection point group has a shielding effect that shields the vehicle from the approaching object, the approaching object is excluded from collision determination; The shielding effect is evaluated using a gap threshold based on a vehicle width, The detection information includes detection information detected by a sonar, The sensor control unit does not exclude the detection point detected by the sonar from the collision determination, The sensor control unit selects detection points to be evaluated for the shielding effect from the detection points detected by the sonar based on the reception strength of the sonar. Vehicle control device.

2. The gap threshold based on the vehicle width is set to a value equal to or smaller than the vehicle width of a passenger vehicle, and the sensor control unit compares the gap of the detection point with the gap threshold, and when the gap of the detection point is wider than the gap threshold, evaluates the shielding effect lower than when the gap of the detection point is narrower than the gap threshold. The vehicle control device according to claim 1.

3. The direction is determined based on one of the direction of the approaching object, the traveling direction of the vehicle, and the traveling direction of the approaching object, The shielding effect is evaluated by evaluating the gaps of the detection points based on the reference direction. The vehicle control device according to claim 2.

4. Gaps between the detection points of the detection point group located in the reference direction are evaluated, and in evaluating the shielding effect, a gap having a small deviation angle with respect to the reference direction is evaluated more heavily than a gap having a large deviation angle with respect to the reference direction. The vehicle control device according to claim 3.

5. Evaluating the number of detection points of the detection point group located in the reference direction, and evaluating the shielding effect higher when the number of detection points is large than when the number of detection points is small. The vehicle control device according to claim 3.

6. When there is a parking target position for parking the vehicle, if the approaching object is in a direction based on the parking target position, the evaluation of the shielding effect is lowered. The vehicle control device according to any one of claims 1 to 5.

7. When there is a parking target position where the vehicle is to be parked, when the detection point cloud is closer than a position based on the parking target position, the shielding effect is evaluated lower than when the detection point cloud is farther from the position based on the parking target position. The vehicle control device according to any one of claims 1 to 6.

8. The detection information includes detection information detected by a radar, and the information on the approaching object is based on the detection information detected by the radar. The vehicle control device according to claim 1.

9. The evaluation target of the shielding effect includes an obstacle detected by the radar. The vehicle control device according to claim 8.

10. the sensor control unit generates an approximation line based on positions of the plurality of detection points included in the detection point group, calculates an average distance from the generated approximation line to positions of the plurality of detection points, or a variance of positions of the plurality of detection points relative to the approximation line, and evaluates the shielding effect as high when the calculated average distance or the calculated variance is small. The vehicle control device according to claim 1.

11. an obstacle detection means for detecting an obstacle around the vehicle and outputting detection information; a sensor control unit that performs a collision determination to evaluate a possibility of a collision with the obstacle, The sensor control unit, based on the detection information, Information on an approaching object that is an obstacle approaching the vehicle; and Generate information on a detection point cloud, which is a set of detection points indicating stationary obstacles; If the detection point group has a shielding effect that shields the vehicle from the approaching object, the approaching object is excluded from collision determination; The shielding effect is evaluated using a gap threshold based on a vehicle width, The detection information includes detection information detected by a sonar, The sensor control unit does not exclude the detection point detected by the sonar from the collision determination, The sensor control unit selects detection points to be evaluated for the shielding effect from the detection points detected by the sonar based on the reception strength of the sonar. vehicle.

12. A vehicle control method executed by one or more computers mounted on a vehicle, comprising: Obtaining detection information of an obstacle detected around the vehicle; Based on the detection information, Information on an approaching object that is an obstacle approaching the vehicle; and Generate information on a detection point cloud, which is a set of detection points indicating stationary obstacles; evaluating a shielding effect of the detection point cloud shielding the vehicle from the approaching object based on information about the approaching object and information about the detection point cloud; excluding the approaching object from a collision determination target for evaluating a possibility of a collision with the vehicle according to the shielding effect; The shielding effect is evaluated using a gap threshold based on a vehicle width, The detection information includes detection information detected by a sonar, The detection point detected by the sonar is not excluded from the collision determination, selecting detection points to be evaluated for the shielding effect from the detection points detected by the sonar based on the reception intensity of the sonar; A vehicle control method.

13. A vehicle control program executed by one or more computers mounted on a vehicle, acquiring detection information of an obstacle detected around the vehicle; Based on the detection information, Information on an approaching object that is an obstacle approaching the vehicle; and A step of generating information of a detection point cloud which is a set of detection points indicating a stationary obstacle; evaluating a shielding effect of the detection point cloud shielding the vehicle from the approaching object based on a gap threshold based on a vehicle width, information about the approaching object, and information about the detection point cloud; excluding the approaching object from a collision determination target for evaluating a possibility of a collision with the vehicle according to the occlusion effect; A vehicle control program, The detection information includes detection information detected by a sonar, The detection point detected by the sonar is not excluded from the collision determination, selecting detection points to be evaluated for the shielding effect from the detection points detected by the sonar based on the reception intensity of the sonar; Vehicle control program.

Images