JP4715579B2 - Potential risk estimation device - Google Patents

Description

  The present invention relates to a potential risk estimation device, and more particularly to a potential risk estimation device that estimates a potential risk level for a jumping pedestrian who jumps out of a barrier ahead of a traveling vehicle to a roadway and presents an appropriate vehicle speed at which the vehicle can travel safely. About.

  Conventionally, as a technology for avoiding collisions with obstacles based on road structure and weather conditions, and supporting safe driving of vehicles, sensor search limit range, blind spot area by surrounding objects, and blind spot shielded by road gradient A region where the unsearchable region obtained from the region overlaps with the estimated travel range of the vehicle is obtained as an excess travel region, and the deceleration or steering angle such as obstacle avoidance is optimized or suitable according to the position and shape of the excess travel region. A driving support system that determines a value (for example, see Patent Document 1) is known.

In addition, the travel block number is identified from the GPS position information, and a travel history database associated with the road environment parameter derived from the road environment real time information and the vehicle speed is constructed. In the playback mode, the travel block number and the road environment parameter are keyed. As a target vehicle speed determination device (for example, refer to Patent Document 2) that searches a travel history database and outputs a suitable target vehicle speed according to the weather and sunshine conditions, and detects a road edge and a vertical line segment from a road image Also known is a driving support device (for example, see Patent Document 3) that identifies a place where the two intersect as a dangerous scene and notifies the driver as a dangerous scene when the moving speed with respect to the road surface of the vertical line is below a predetermined value. Yes.
JP 2001-34898 A JP 2003-276472 A JP 2001-252550 A

  However, since it has been difficult to detect sudden jumping pedestrians using conventional techniques, a support system for preventing jumping pedestrian accidents using sensing has not been proposed.

  In the technique of Patent Document 1, since only the collision with a stationary obstacle existing in the blind spot on the course of the own vehicle is considered, the collision accident with a moving object such as a pedestrian jumping out from the road side to the roadway can be reduced. Therefore, there is a problem in that it is impossible to estimate the degree of risk and to set an appropriate vehicle speed.

  In the technique of Patent Document 2, a suitable target vehicle speed is automatically output from past travel history data according to the current vehicle travel position, weather, sunshine conditions, etc. Appropriate vehicle speed that takes into account structural changes cannot be set appropriately. Furthermore, there is a problem that the target vehicle speed cannot be set for a road having no travel history.

  In the technique of Patent Document 3, the blind spot is detected as a vertical line segment intersecting the road edge on the image and identified as a dangerous scene, and a warning is given, but a method for setting an appropriate vehicle speed is not shown. In addition, since the existence of a barrier such as a guardrail that separates the pedestrian from the host vehicle is not taken into account, there is a problem in that the barrier and the blind spot are misrecognized, or an excessive warning is given for the blind spot.

  The present invention has been made to solve the above-described problems, and estimates the potential risk level for a jumping pedestrian in front of the vehicle, and teaches and warns the driver of an appropriate vehicle speed at which the vehicle can safely travel. An object is to provide an apparatus.

  In order to achieve the above object, a latent risk estimation device according to claim 1 is characterized in that at least one of a blind spot in a driving environment based on a driver's viewpoint and a blind spot in a driving environment based on a sensor mounting position is used. Based on the blind spot detection means to detect, the barrier detection means to detect the roadside object that prevents the pedestrian from entering the roadway as a barrier, the blind spot detected by the blind spot detection means and the barrier detected by the barrier detection means Then, when there is a jumping pedestrian jumping out from the barrier to the roadway, a position where the pedestrian can be found is estimated first, and a collision probability between the own vehicle and the pedestrian is estimated based on the estimated position. A collision probability estimating means.

  According to the first aspect of the present invention, not only a blind spot caused by a change in the road structure and visual field, but also a roadside object that prevents a pedestrian from entering the roadway is detected as a barrier, and a steep jumping walk that is difficult to detect with a sensor. The collision probability for the person can be estimated.

  The latent risk estimation device according to claim 2 is the latent risk estimation device according to claim 1, wherein the blind spot is caused by a blind spot caused by a road structure, a blind spot caused by a shield, and a change in a visual field of a driver or a sensor. It is composed of at least one of blind spots.

  According to the second aspect of the present invention, the blind spot is detected from a plurality of aspects including a blind spot caused by the positional relationship between the vehicle position and the road structure, a blind spot caused by the shielding object, and a blind spot caused by visual field variation caused by weather and lighting conditions. can do.

  The latent risk estimation device according to claim 3 is the latent risk estimation device according to claim 1 or 2, wherein the occurrence probability of a jumping pedestrian is calculated based on time, map information, weather, and traffic information. The collision probability estimation means further estimates the collision probability based on the estimation result by the probability fluctuation estimation means.

  According to the third aspect of the present invention, the fluctuation amount of the pedestrian occurrence probability is estimated in consideration of the pedestrian crossing needs that fluctuate in accordance with the change in the traffic volume according to the surrounding facility and the time zone of the own vehicle position. Based on this, the occurrence probability of a jumping pedestrian is corrected, so that a more realistic collision probability can be estimated.

  The latent risk estimation apparatus according to claim 4 is the latent risk estimation apparatus according to any one of claims 1 to 3, further comprising an appropriate vehicle speed setting means for setting an appropriate vehicle speed based on the collision probability. I have.

  According to the invention described in claim 4, safe driving can be promoted by setting an appropriate vehicle speed based on the estimated collision probability and presenting it to the driver.

The blind spot detecting means includes a driver-side blind spot detecting means for detecting a blind spot based on a driver's viewpoint, and a sensor-side blind spot detecting means for detecting a blind spot based on a sensor mounting position. The probability estimation means estimates the driver-side collision probability based on the blind spot detected by the driver-side blind spot detection means and the barrier detected by the barrier detection means, and is detected by the sensor-side blind spot detection means. Based on the blind spot and the barrier detected by the barrier detection means, the collision probability on the sensor side is estimated, and based on the comparison result between the estimated collision probability on the driver side and the collision probability on the sensor side, walking with the own vehicle Estimate the collision probability with the person.

This makes it possible to estimate the driver's potential risk based on the sensor. Moreover, the troublesomeness of the driver can be reduced by setting the appropriate vehicle speed and performing the warning only when the collision probability on the driver side is higher than the collision probability on the sensor side.

  As described above, according to the present invention, focusing on the physical conditions of the driving environment due to blind spots and barriers, the risk of collision with respect to a sudden pedestrian jump that is difficult to detect with a sensor is set to a certain value or less. The effect of being able to encourage driving safely at an appropriate vehicle speed that can be maintained is obtained. In addition, since it is possible to prevent the driver from excessively reducing the vehicle speed due to excessive safety, the effect of suppressing an increase in travel time and preventing the flow of traffic can be obtained.

  Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows a configuration diagram of a latent risk estimation apparatus 10 according to the present embodiment. As shown in the figure, the latent risk estimation device 10 includes a sensor group for detecting a traveling state and an external environment state of the host vehicle, a map database 16 storing map information, and detection data from these sensor groups. A control device 11 that controls the latent risk estimation device 10 based on the map information is provided.

  As a sensor for detecting the traveling state of the host vehicle, a vehicle speed sensor 12 for detecting the vehicle speed is provided. In addition, as a sensor for detecting an external environmental state, a camera 13 for photographing the front of the host vehicle, a laser radar 14 for detecting a three-dimensional object (obstacle) in front of the host vehicle, and a GPS device 15 for detecting the position of the host vehicle. , And a weather sensor 17 including a rain sensor, a temperature sensor, a humidity sensor, and the like. A millimeter wave radar may be provided instead of the laser radar 14 or together with the laser radar 14.

  The camera 13 is composed of a small CCD camera or CMOS camera, and is attached to the upper part of the front window of the vehicle so as to photograph the front of the vehicle. Image data such as the road conditions ahead taken by the camera 13 is input to the control device 11 constituted by a microcomputer or the like. By using a far infrared camera as the camera 13, a pedestrian as an obstacle can be detected more reliably.

  The laser radar 14 receives a near-infrared light pulse reflected from a light emitting element composed of a semiconductor laser that emits near-infrared pulses and scans in the horizontal direction, and obstacles ahead (roadside objects, surrounding vehicles, etc.). And a light receiving element, and is attached to the front grille or bumper of the vehicle. The distance measurement principle by the laser radar 14 is obtained by measuring the arrival time of the reflected light pulse until it is received by the light receiving element with reference to the time point when the light is emitted from the light emitting element, and obtaining it by the product of the speed of light. By horizontally scanning the laser beam with the scanning mechanism, the shape of the obstacle ahead of the vehicle can be measured. The measurement data for each scan is obtained as continuous data of the scan angle, the measurement distance, and the reflection intensity of the laser beam, and is input to the control device 11.

  FIG. 3 is a polar coordinate display of the scan angle and measurement distance of the laser radar 14 with the vehicle position as the origin, and FIG. 4 is a known coordinate in an orthogonal coordinate system obtained by looking down on the road from above. It is converted by the conversion means and displayed in orthogonal coordinates. In the orthogonal coordinate system of FIG. 4, the z-axis (vertical axis) is the distance in the vehicle traveling direction, and the x-axis (horizontal axis) is the distance in the pedestrian crossing direction.

  From the GPS device 15, position data indicating the position of the host vehicle is output and input to the control device 11. The control device 11 reads map information including road shape data and the like stored in the map database 16 and displays the position of the host vehicle on the map represented by the read map information. Since the map information includes road shape data, the position of the detected obstacle on the road can be obtained.

  Next, the control routine of the latent risk estimation apparatus in the present embodiment will be described with reference to the flowchart shown in FIG. This control routine is stored in the memory of the control device 11 of the potential risk estimation device.

  First, in step 100, data detected by a sensor that detects the traveling state of the host vehicle and the external environmental state is captured. Specifically, scan data from the laser radar 14, own vehicle position data from the GPS device 15, vehicle speed data from the vehicle speed sensor 12, and the like are taken into the control device 11. As will be described later, the blind spot and the barrier can be detected from the scanned data acquired here, but for the sake of simplicity of explanation, it is assumed that the viewpoint of the driver and the origin of the laser radar 14 coincide.

  Next, in step 110, the acquired scan data of the laser radar 14 is grouped. In grouping, the scan data is plotted against the scan angle versus distance, and a region where the distance is continuous is set as a single obstacle candidate. However, since some scan data is missing, if the distance between adjacent scan data is 30 cm or less in the orthogonal coordinate system with reference to the thickness of the utility pole of about 30 cm, they are combined and regarded as the same obstacle.

  Obstacles around the host vehicle include traveling vehicles (preceding vehicles, oncoming vehicles, two-wheeled vehicles, etc.), moving objects such as pedestrians, and parked vehicles, roadside objects (electric poles, There are stationary objects such as guardrails, etc., and block structures (buildings, houses, fences, etc.). In the present embodiment, the detected obstacle is classified into a moving object and a stationary object. In this case, a moving object having a size of about 0.4 m and crossing the road at a speed of about 1.2 m / s in the transverse direction is detected as a pedestrian candidate.

  In step 120, a blind spot boundary caused by the shielding of the road structure or obstacles is detected. A portion where the distance discontinuously changes by plotting the scan angle versus distance of the laser radar 14 can be regarded as an area where obstacles overlap each other. The back of the object in front is not visible directly from the vehicle, so it becomes a blind spot. That is, a portion where the distance changes discontinuously is detected as a boundary of a blind spot caused by road structure or obstacle shielding.

  Alternatively, a road structure registered in the map database 16 may be used to detect an area where the line of sight cannot be seen from the positional relationship between the vehicle position and the road structure (road shape) by the GPS device 15 as a blind spot. Areas where the line of sight is not good include the tip of a curve and an intersection.

  In step 130, a blind spot resulting from the visual field is detected. The field of view is defined by the viewing distance and the viewing angle, and varies mainly depending on lighting conditions and weather conditions.

  First, the viewing distance will be described. As for the lighting conditions, the lighting conditions of the driving environment are determined by a headlight switch, and the viewing distance is set to, for example, 100 m in the daytime, and the viewing distance is set to a shorter distance than in the daytime, for example, 40 m in the case of the passing beam at night.

  With regard to the weather conditions, it is determined whether or not it is raining with a wiper switch, and in the case of raining, the setting value of the visual distance is switched to 200 m or less. Further, the fog lamp switch determines the presence or absence of fog, and in the case of fog, the setting value of the visual distance is switched to 50 m or less. The weather condition may be determined by a clear rain sensor of the weather sensor 17.

  Regarding the viewing distance of the driver, when both the laser radar 14 and the camera 13 are mounted on the vehicle, the viewing distance can be estimated by comparing the both. If the position coordinates where the laser radar 14 and the camera 13 are mounted are known by calibration, the viewing distance can be estimated by comparing the image processing with the laser radar 14. The scan data can be projected onto the image by the existing coordinate conversion means, and it can be expected that the edge of the object exists at the end point of the scan data grouping. The maximum distance at which the edge line of the object can be detected by the image processing edge extraction means is estimated as the viewing distance. Similarly, when both the millimeter wave radar and the camera 13 are mounted on the vehicle, the viewing distance can be estimated by comparing the both.

  Next, the viewing angle will be described. Since the viewing angle differs between the driver and the sensor, each case will be described.

  As for the driver, since the peripheral visual field is wide and sensitive to moving objects, only the viewing distance is considered without limiting the viewing angle.

  As a sensor, a laser radar 14 or a camera 13 is used. However, there is a physical limit to the viewing angle of the detection range, and anything existing outside the viewing angle cannot be detected. Therefore, the boundary of the visual field is preset as a blind spot boundary.

  In step 140, a barrier is detected. A barrier prevents pedestrians from entering the vehicle lane. Due to the presence of the barrier, the probability of sudden jumping out of pedestrians and bicycles is reduced.

  When the data of the scan type laser radar 14 attached in front of the host vehicle 20 is plotted against the scan angle versus distance, it can be considered that an obstacle exists in a portion where the distance data is spatially continuous. A continuous obstacle present on the roadside is detected as a barrier 22 that prevents a pedestrian from entering the roadway. Typical barriers 22 are the parked vehicle shown in FIG. 5 and the guardrail shown in FIG.

  Specifically, the grouped scan data is projected onto an orthogonal coordinate system, and the length of the data point sequence arranged in the traveling direction is, for example, about 10 m or more, and the angle change between the end points of the point sequence is, for example, 45 ° or less. Are determined to be guardrail candidates.

  As another method for determining the guard rail, the guard rail is composed of struts arranged at equal intervals and pipes connecting the struts. Therefore, the magnitude of the reflection intensity of the point sequence is periodically (for example, approximately every 2 m) according to the interval of the struts. Those that change to are determined as guardrail candidates. Alternatively, for example, a point sequence of 2 m intervals aligned in the road extension direction is determined as a guardrail candidate.

  From obstacles other than the guardrail candidates determined as described above, for example, a point sequence having a length of 1.5 m or more is set as a vehicle candidate (including a motorcycle).

  Although the guard rail composed of thin pipes is difficult to detect with the laser radar 14, the detection rate is improved by widening the beam divergence angle of the laser light.

  Although the scanning laser radar has been described in the present embodiment, the blind spot and the barrier can be similarly detected by stereo image processing and two-dimensional scanning laser radar. In this case, obstacle height information can be obtained by grouping obstacle candidates and statistically obtaining the distribution in the height direction. In Japan, guardrails are standardized to a height of about 80 cm, so it is possible to discriminate guardrail candidates from height information.

  Also, since the utility poles / signs and guardrails are fixed on the roadside for a long period of time, the guardrail positions registered in the map are registered in advance, and the blind spots and barrier positions are determined by the own vehicle position obtained by the positioning means such as the GPS device 15. May be estimated.

In step 150, the collision probability with a jumping pedestrian is estimated. Here we assume the following.
(A) The probability that a jumping pedestrian appears on the roadway is constant per unit distance and unit time in a certain section.
(B) The jumping pedestrian crosses the roadway at a constant speed without noticing the vehicle.
(C) The driver is excellent and all pedestrians who are not hiding in the blind spot are aware. In addition, if a pedestrian jumping out of the blind spot is found, he immediately acts.

  Based on the above assumption, the collision probability is calculated as a physical collision. At this time, the collision probability of a jumping pedestrian in front of the vehicle depends on the relative position and relative speed at the time when the driver first finds the pedestrian and applies braking.

Therefore, the collision probability is calculated in the following two steps.
(1) When a pedestrian jumps out from the blind spot / barrier arrangement in front of the vehicle, the position where the driver first finds the pedestrian is estimated.
(2) The collision probability is estimated by integrating the occurrence probability of the pedestrian jumping out from the position.

  Here, the following variables are defined.

v: vehicle speed vc: vehicle during impact speed z _1: Relative distance (vehicle traveling direction) between the vehicle and the pedestrian at the time of jumping out the pedestrian
zb: braking stop distance (= v ² / (2 · ac))
x ₁ : Relative position when pedestrian jumps out (pedestrian crossing direction)
vp: Jumping pedestrian crossing speed (generally 1.2 m / s is set)
tc: Time when the vehicle reaches the distance of the pedestrian (time when the vehicle passes or collides with the pedestrian)
ac: Vehicle deceleration (assuming emergency braking, set at 0.6 m / s ² )
e: Relative position of vehicle and pedestrian at time of collision tc: Horizontal width of own vehicle pc: Probability of jumping pedestrian per unit time and unit distance (in this embodiment, pc is a constant value in a certain section (Normally, the probability value is 1 or less, but it is calculated with pc = 1 for convenience of calculation.)
FIG. 7 shows the positional relationship between the host vehicle 20 and the pedestrian 24 at the start of braking, and FIG. 8 shows the positional relationship when the host vehicle 20 reaches the position of the pedestrian 24.

  At this time, the relative position e of the host vehicle 20 and the pedestrian 24 at the time tc can be calculated by the equation (1), and the host vehicle speed vc can be calculated by the equation (2).

At time tc, the situation where the front of the vehicle and the pedestrian overlap, that is,
-w / 2 <e <w / 2
A collision occurs when the condition is satisfied. Here, it is assumed that the size of the pedestrian is negligible compared to the vehicle. When calculating strictly, a pedestrian width should just be added to w.

  The lateral position xmin where the pedestrian reaches -w / 2 at time tc and the lateral position xmax where the pedestrian reaches w / 2 can be calculated by Equations (3) and (4).

  By arranging the above equations, equations (5) and (6) are obtained.

  Further, the braking stop distance zb can be calculated by Expression (7).

  As shown in FIG. 9, the collision with the pedestrian jumping out from the lateral position xmin to xmax cannot be avoided even if the distance is equal to or less than the braking stop distance zb.

  As an example, a collision probability calculation in the case of a monotonous blind spot where x is constant in which electric poles, signs, trees, gates, and the like are continuously generated as shown in FIGS. As for the current collision probability Pb of the host vehicle, the jumping pedestrian occurrence probability pc may be integrated in the z-direction range of the blind spot region zmin to zmax surrounded by xmin to xmax as shown in the equation (8).

  Here, Lb is the distance in the z direction of the blind spot surrounded by the lines xmin and xmax.

  Next, a method of calculating the collision probability Pb when there are both blind spots and barriers (parked vehicles) as shown in FIG. Here, it is assumed that no pedestrian jumps out of the barrier or the probability is very low.

  Excluding the barrier, the total sum Lb of the distances in the traveling direction z direction of the blind spot area surrounded by the lines of the blind spot xmin and xmax is calculated. Further, when the barrier is on the side of the roadway from the blind spot when viewed in the transverse direction, it is regarded as a barrier. At this time, the collision probability can be calculated by Equation (9).

  The collision probability when considering the visual field (sight distance) is calculated as follows.

  Consider a condition in which the crossing pedestrian 24 suddenly becomes visible from a visual range zobs as shown in FIG. In the case of the jumping pedestrian occurrence probability pc, the crossing pedestrian occurrence probability po on the zobs line viewed from the host vehicle 20 is expressed by Expression (10).

  If zobs is less than the braking stop distance zb, a collision occurs. zb is obtained by the above-described equation (7).

  From the above assumption, the jumping pedestrian moves at a constant speed on the zobs line. Therefore, if po is integrated in the range of the vehicle width w, the collision probability Pobs according to the visual distance can be calculated by the equation (11).

  When considering the blind spot, the barrier, and the visual distance, the collision probability P is calculated by Expression (12).

  In step 160, the fluctuation amount of the occurrence probability of the jumping pedestrian is estimated. In the present embodiment, it has been assumed that the occurrence pedestrian occurrence probability pc per unit time and unit distance is a constant value within a certain area, but there are places where pedestrians need to cross the road. Therefore, the frequency of the occurrence of the jumping pedestrian, that is, the fluctuation amount of the occurrence probability of the jumping pedestrian is estimated from the surrounding map information, and the value of the jumping pedestrian occurrence probability pc is set.

  For example, if there is a nursery school, a school, a station, a bus terminal, a shopping street, etc., the peripheral facilities registered in the map database 16 are searched based on the vehicle position obtained from the GPS device 15, and the jumping pedestrian occurrence probability pc is detected. Set a larger value.

  In addition, a large value of the jumping pedestrian occurrence probability pc is set in the commuting time zone from 7:00 am to 9:00 am and from 5:00 pm to 7:00 pm.

  Furthermore, since it is difficult for pedestrians to cross a road with a heavy traffic volume, the value of the pedestrian occurrence probability pc may be reduced based on the VICS information.

  The collision probability is corrected based on the jumping pedestrian occurrence probability pc set in this way.

  In the present embodiment, the collision probability calculation method based on the integral calculation is shown assuming a relatively simple situation. However, the collision probability may be calculated by the Monte Carlo method by simulating the pedestrian jumping on the computer. .

  In step 170, an appropriate vehicle speed for keeping the collision probability below the reference value is set.

  The collision probability with respect to the virtual jumping fluctuates due to a change in the blind spot or barrier in front of the vehicle. Moreover, the collision probability varies depending on the vehicle speed, and the collision probability decreases as the vehicle speed decreases.

  Here, it is assumed that a good driver is driving at the maximum vehicle speed at which the collision probability is below a certain value. Since the relationship between the collision probability P and the vehicle speed v is known from the above-described calculation formula, the vehicle speed that keeps the collision probability below a certain value can be set.

  As the reference value of the collision probability, the collision probability P at the time of steady running is stored in a memory (not shown), or the average value of the collision probability P in a past fixed period may be used as the reference value. .

  The collision probability P when the vehicle speed v is changed is recalculated from the blind spot and the barrier detected earlier. However, considering the resolution of the vehicle speed presented to the driver, the vehicle speed is practically 60 km / h and 40 km / h. There are at most several stages, such as 20 km / h. Therefore, it is sufficient to recalculate at each vehicle speed, and the calculation cost does not increase. And the maximum value of the recalculation value of the collision probability which becomes below the reference value is set as the appropriate vehicle speed.

  In setting the appropriate vehicle speed, the appropriate vehicle speed may be set so that the value obtained by adding the weight to the collision probability and correcting it is kept at the reference value. The weight of the pedestrian can be considered as the weight. In particular, in a pedestrian accident, the degree of obstacles that a pedestrian receives depends on the vehicle speed, and the lower the speed, the less the damage. Therefore, simply, the term of the square of the vehicle speed proportional to the kinetic energy of the vehicle may be added to the collision probability. The vehicle speed vc at the time of a collision with a pedestrian jumping out at a forward distance z can be calculated from equation (2).

  Therefore, the weighted collision probability D can be calculated by Expression (13) or Expression (14).

  The weight based on the damage degree may be set by statistically investigating the relationship between the vehicle speed and the damage at the time of collision from the actual accident statistical data.

In step 180, the set appropriate vehicle speed is presented to the driver. This can encourage the driver to drive safely. The method of presenting the appropriate vehicle speed to the driver differs depending on the application, but the following methods are conceivable.
(A) The appropriate vehicle speed is displayed on the meter.
(B) Encourage the driver to accelerate or decelerate by voice.
(C) An alarm is issued when the current vehicle speed is faster than the appropriate vehicle speed.
(D) When the current vehicle speed is faster than the appropriate vehicle speed, intervention braking is performed at the appropriate vehicle speed.

  When presenting the appropriate vehicle speed, the collision probability on the driver side estimated based on the blind spot and the barrier based on the driver's viewpoint, and the sensor side estimated based on the blind spot and the barrier based on the sensor mounting position are used. The collision probability can be compared and presented only when the collision probability on the driver side is large. With such a configuration, the troublesomeness of the driver can be reduced.

  After presenting the appropriate vehicle speed in this way, the process returns to step 100 and the above-described processing is repeated.

  As described above, in the present embodiment, the driver response delay time is not taken into consideration for the sake of simplification. However, it is also possible to take into account the detailed case classification of conditions.

  In the present embodiment, the straight single road has been described. However, the calculation can be similarly performed on a curved road. In the case of a curved road, equations (5) and (6) may be corrected based on curvature information. The curvature information can be obtained from the map database 16 or the result of detection of the road by the laser radar 14.

  On a highly safe road where pedestrian separation has progressed due to a guardrail or the like, the appropriate vehicle speed may be estimated to be higher than the legal speed (60 km / h). In this case, the legal speed is given priority. In Japan, the Public Safety Commission may set speed limits for individual roads, but no speed exceeding this limit will be set.

It is a block diagram which shows the latent risk estimation apparatus which concerns on this invention. It is a flowchart which shows the flow of control of the latent risk estimation apparatus which concerns on this invention. This is laser radar scan data (polar coordinate display). This is laser radar scan data (rectangular coordinate display). It is a figure which shows the example of the barrier and blind spot by a parked vehicle. It is a figure which shows the example of the barrier by a guardrail. It is a figure which shows the positional relationship of the vehicle and pedestrian at the time of a braking start. It is a figure which shows the positional relationship when a vehicle arrives at a pedestrian position. It is a figure which shows the appearance range of the jumping out pedestrian who cannot avoid a collision. It is a figure which shows the example of a monotone blind spot. It is a figure which shows the collision in a monotone blind spot. It is a figure which shows the collision by the blind spot and barrier of a parked vehicle. It is a figure which shows the visual field at the time of a night passing beam.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Potential risk estimation apparatus 11 Control apparatus 12 Vehicle speed sensor 13 Camera 14 Laser radar 15 GPS apparatus 16 Map database 17 Weather sensor 20 Own vehicle 22 Barrier 24 Pedestrian

Claims (4)

A blind spot detecting means for detecting at least one of a blind spot in the driving environment based on the viewpoint of the driver and a blind spot in the driving environment based on the mounting position of the sensor;
Barrier detection means for detecting a roadside object that prevents a pedestrian from entering the roadway as a barrier;
Based on the blind spot detected by the blind spot detection means and the barrier detected by the barrier detection means, when there is a jumping pedestrian jumping out from the barrier to the roadway, a position where the pedestrian can be found first is estimated. A collision probability estimating means for estimating a collision probability between the own vehicle and the pedestrian based on the estimated position;
A potential risk estimation device comprising:   The latent risk estimation device according to claim 1, wherein the blind spot is at least one of a blind spot caused by a road structure, a blind spot caused by a shielding object, and a blind spot caused by a change in the field of view of a driver or a sensor. Probability fluctuation estimation means for estimating the fluctuation of the occurrence probability of a jumping pedestrian based on time, map information, weather, and traffic information,
3. The potential risk estimation apparatus according to claim 1, wherein the collision probability estimation unit corrects the collision probability based on an estimation result by the probability variation estimation unit.   The latent risk estimation device according to any one of claims 1 to 3, further comprising appropriate vehicle speed setting means for setting an appropriate vehicle speed based on the collision probability.