JP7519028B2 - Vehicle control device - Google Patents

Description

The present invention relates to a vehicle control device that performs distance correction (distance determination) using a probabilistic model when an abnormality occurs in distance detection by an on-board radar.

2. Description of the Related Art Recently, it has become common for vehicles to be equipped with radar for controlling various in-vehicle devices, such as automatic brake control, adaptive cruise control, and headlamp illumination range control.

Patent Literature 1 discloses a system for determining a target object using an on-board radar and an on-board camera. Patent Literature 2 discloses a system for estimating a reflectance distribution of a radar laser using an on-board camera and changing a reception intensity threshold.

JP 2007-304033 A JP 2013-019684 A

However, when the distance to a target vehicle recognized by an on-board camera is detected by an on-board radar, an abnormality may occur in the distance detection due to various noises, etc. Abnormal states of distance detection include, for example, a case where distance detection itself is impossible, or a detected distance is abnormally large or abnormally small. If an abnormality occurs in the distance detection by the on-board radar, a malfunction will occur in the control of various on-board control devices that are executed according to the detected distance.

The present invention has been made in consideration of the above-mentioned circumstances, and its purpose is to provide a vehicle control device that can perform alternative distance correction (distance determination) by using a probabilistic model when an abnormality occurs in distance detection by an on-board radar.

The present invention is based on the fact that the delay time when a detection element wave consisting of radio waves or light is sent back and forth between the antenna position and a point on a target vehicle is constant regardless of the radar type, and that distance is calculated using the delay time as a variable regardless of the radar type.

The probability model used in the present invention is data stored by simulating a number of delay times, which are the round trip times of a number of detection elementary waves between a virtual vehicle and a virtual antenna whose position is changed over a number of locations in a predetermined range while taking into account multipath, and the number of different delay times is associated with each of the different delay times . A number of such probability models are prepared according to combinations of differences in the relative positional relationship between the virtual antenna and the virtual vehicle and differences in the shape of the virtual vehicle. When an abnormality occurs in the distance detection by the on-board radar, a specific probability model according to the relative positional relationship with the target vehicle recognized by the on-board camera and the shape of the target vehicle is selected, and the distance corresponding to the delay time stored in this specific probability model is determined as the corrected distance.

Specifically, the present invention employs the following solution.

An onboard camera that captures images of the surrounding environment and recognizes target vehicles;
an on-vehicle radar for detecting a distance to the target vehicle;
a plurality of probability models that are created in advance by simulation in accordance with a plurality of delay times, which are the round trip times of a plurality of detection element waves between a virtual vehicle and a virtual antenna whose position is changed over a plurality of locations within a predetermined range, and that correspond to and store the number of acquired delay times for each different delay time, and that are created in accordance with a plurality of relative positional relationships between the virtual antenna and the virtual vehicle and differences in the shape of the virtual vehicle;
an abnormality detection means for detecting an abnormality in distance detection by the on-vehicle radar;
a selection means for selecting, when an abnormality is detected by the abnormality detection means, from among the multiple probability models, a specific probability model corresponding to a relative positional relationship with the target vehicle and a shape of the target vehicle acquired by the vehicle-mounted camera;
a correction means for correcting a distance to the target vehicle based on a delay time stored in the specific probability model when an abnormality is detected by the abnormality detection means;
It is designed to be equipped with the following.

According to the above-mentioned solution, the specific probability model indicates the distribution of delay times (i.e., distances according to the delay times) of the detection element waves of the on-board radar in response to the relative positional relationship between the vehicle and the target vehicle and the shape of the target vehicle, and corresponds to a distance in a range that is highly likely to be the actual distance. Therefore, when an abnormality occurs in the distance detection of the on-board radar, the distance to the target vehicle can be corrected (determined) to a highly likely distance range based on the specific probability model. In particular, since the probability model is created taking into account that the antenna position moves according to the movement of the vehicle, causing fluctuations in the detected distance, it is even more preferable for the distance determined by correction to be in a highly likely distance range.

Based on the above solution, a preferred embodiment such as the following can be adopted.

When the abnormality detection means detects an abnormal state in which the on-board radar cannot detect the distance, the correction means can determine the distance corresponding to the delay time selected from the delay times stored in the specific probability model as the corrected distance. In this case, the distance can be determined when the on-board radar cannot detect the distance.

The correction means may determine the distance after correction to be the distance corresponding to the delay time in the vicinity of the delay time having the largest number of acquisitions among the delay times stored in the specific probability model. In this case, the corrected distance may be set to be the distance with the highest probability.

When the abnormality detection means detects an abnormal state in which the detection distance by the on-board radar is abnormally large, the correction means can determine the distance after correction to be the maximum delay time or a distance corresponding to the vicinity of the maximum delay time among the delay times stored in the specific probability model. In this case, the corrected distance can be determined without shortening the distance unnecessarily.

When the abnormality detection means detects an abnormal state in which the detection distance by the on-board radar is abnormally too small, the correction means can determine the distance after correction to be the minimum delay time among the delay times stored in the specific probability model or a distance corresponding to the vicinity of the minimum delay time. In this case, the corrected distance can be determined without increasing the distance unnecessarily.

The predetermined range may be any one of the ranges in the forward/backward direction, the left/right direction, and the up/down direction of the virtual antenna, a range consisting of any two of them, or a range consisting of all of them. In this case, it is preferable to set the correction distance as a likely distance range as a probability model corresponding to the movement direction of the antenna position that affects the distance detection.

The multiple relative positional relationships set when creating the multiple probability models may be a combination of the azimuth of the virtual vehicle with respect to the virtual antenna, the direction of the virtual vehicle, and the distance from the virtual antenna to the virtual vehicle. In this case, it is preferable to prepare multiple appropriate probability models according to various situations, and to select a specific probability model that matches the current situation and set the correction distance as within a highly probable distance range.

According to the present invention, when an abnormality occurs in the distance detection by the vehicle-mounted radar, a probability model can be used to perform alternative distance correction (distance determination).

FIG. 2 is a diagram showing a situation in which a radar mounted on the vehicle detects the distance to a vehicle ahead. 13A and 13B are diagrams showing a situation in which a simulation is being performed using detected elementary waves and a predetermined range in which the virtual antenna position is changed. FIG. 13 is a diagram showing an example of distances according to delay times obtained at multiple positions within a predetermined range. FIG. 13 is a diagram showing an example of a probability model. 11 is a flowchart showing an example of control for obtaining a probabilistic model. 4 is a flowchart for obtaining the characteristics of an on-board radar to be verified using a probabilistic model. A simplified diagram showing a situation in which distance correction is performed using a probability model. FIG. 2 is a diagram showing an example of a control system of the present invention. 4 is a main flow chart showing a control example of the present invention. 5 is a flowchart showing an example of distance correction in a follow-up constant speed control device. 5 is a flowchart showing an example of distance correction in an automatic braking device. 5 is a flowchart showing an example of distance correction in high/low headlamp switching control.

First, an example of distance detection by an on-board radar will be briefly described with reference to FIG. 1. In FIG. 1, the host vehicle is indicated by V1, and the vehicle ahead of it is indicated by V2. The host vehicle V1 is equipped with an on-board camera S1 that captures an image of the front and recognizes the vehicle ahead of it V2 as a target vehicle. The host vehicle V1 is also equipped with an on-board radar S2, and its antenna is indicated by 1. A detection wave transmitted (irradiated) from the antenna 1 is reflected by the vehicle ahead of it V2 and received by the antenna 1. The distance to the vehicle ahead of it V2 is calculated by substituting the delay time from transmission to reception of the detection wave by the antenna 1 into a predetermined calculation formula set in the on-board radar. If the delay time is the same, the calculated distance is the same regardless of the radar method or the internal model.

If there is some abnormality in the distance detection by the vehicle-mounted radar S2, a probability model, which will be described later, is used to perform distance correction (distance determination).

Next, a simulation for obtaining the above-mentioned probability model will be described with reference to FIGS.

2 shows an antenna FA (corresponding to antenna 1 in FIG. 1) and a virtual vehicle FV (corresponding to forward vehicle V2 in FIG. 1) set in the simulation. The virtual vehicle FV is maintained at a certain fixed position.

Taking multipath into consideration, the virtual antenna FA is changed in position many times within a predetermined range P. In the embodiment shown in Fig. 2, the predetermined range P is changed in position within a two-dimensional plane extending horizontally in the front-rear direction, which is the X-axis direction, and the left-right direction, which is the Y-axis direction, with the origin position (corresponding to the position of the antenna 1 in Fig. 1) as the center. The X-axis direction corresponds to the front-rear direction of the host vehicle V1 passing through the antenna 1 in Fig. 1. The Y-axis direction corresponds to the left-right direction (vehicle width direction) of the host vehicle V1 passing through the antenna 1 in Fig. 1.

The predetermined range P is set to a size of about 50 cm square, with a length of about 25 cm forward and backward from the origin position and a length of about 25 cm to the left and right from the origin position. The range of about 50 cm in the X-axis direction is based on the assumption that the vehicle will move forward and backward due to acceleration/deceleration, etc. The range of about 50 cm in the Y-axis direction is based on the assumption that the vehicle will move about 25 cm left and right within the same driving lane.

The locations to be changed within the predetermined range P are an extremely large number of locations (e.g., about 3000 locations) set in a vertical and horizontal grid pattern for each predetermined unit. In particular, when the vehicle-mounted radar to be verified is a millimeter wave radar, it is preferable to change the location in units of half the wavelength.

2 shows a state in which the virtual antenna FA is located at the origin position, which is the center of the predetermined range P. In the simulation, a detection element wave (one detection wave) is transmitted from the virtual antenna FA located at the origin position toward the virtual vehicle FB. The transmitted detection element wave is reflected from the virtual vehicle FV and received by the virtual antenna FA. The time required from transmission to reception of this detection element wave is the delay time.

The detection element waves used in the simulation are light rays that are not easily affected by noise such as interference. More specifically, in the embodiment, the simulation is performed by a ray tracing method that considers radio waves as light and analyzes the main propagation path of light.

The virtual antenna FA sequentially transmits detection element waves while changing the angle little by little, and scans the virtual vehicle FV. Of the detection element waves irradiated to the virtual vehicle FV, the leftmost detection element wave is indicated by the symbol β1, and the rightmost detection element wave is indicated by the symbol β2. There are many detection element waves between the detection element waves β1 and β2, some of which are indicated by dashed lines in FIG. 3. A delay time is acquired for each detection element wave. Then, the acquired delay times and their number are stored in correspondence with each other, and this stored data is used as a probability model . In the embodiment, the delay times used to create the probability model are limited to the case where the reception strength is equal to or greater than a predetermined value set in advance, but it is also possible to create a probability model using all the acquired delay times.

After the scanning of the virtual vehicle FV at the origin position is completed, the position of the virtual antenna FA is changed, and the delay times are obtained for each of the many detected elementary waves in the same manner as described above, and the delay times are associated with the number of delay times and stored. In this manner, the virtual antenna FA is moved to all positions set within the predetermined range P, and the number of delay times obtained at all positions is counted, for example, in stages for each unit time. Based on this counting, a probability model showing the number of delay times obtained for each different delay time is statistically created.

3 is a diagram showing an example of the delay time (corresponding distance) obtained for all positions in the predetermined range P. It can be seen that even if the position of the virtual antenna FA is changed within the narrow predetermined range P, a large difference occurs in the obtained delay time.

The data, which is stored based on the acquired number of delay times and corresponds to each different delay time, is shown as a probabilistic model in FIG. 4. In the probabilistic model shown in FIG. 4, the delay time is divided into stages for each predetermined unit time, but it can also be continuously variable. In the probabilistic model shown in FIG. 4, the minimum delay time is set as the lower limit (minimum delay time) and the maximum delay time is set as the upper limit (maximum delay time). Note that the probabilistic model in FIG. 4 is set when the shortest distance between the virtual antenna FA at the origin position and the virtual vehicle FV is set to 10 m.

Here, until the minimum delay time and the maximum delay time converge to a certain time, it is required to change the position of the virtual antenna FA at an extremely large number of points within a predetermined range P. To determine how many points the position of the virtual antenna FA needs to be changed (hereinafter referred to as the "number of samples"), the number of samples was changed in various ways, and the minimum delay time and the maximum delay time were obtained for each sample.

First, when the number of samples was set to 5 (a total of 5 locations including the origin position and the four corners), the minimum distance corresponding to the minimum delay time was 9.81 m, and the maximum distance corresponding to the maximum delay time was 13.75 m. When the number of samples was increased to include these five locations, the minimum distance corresponding to the minimum delay time and the maximum distance corresponding to the maximum delay time obtained were as follows:

When the sample size was 100, the minimum distance was 9.74 m and the maximum distance was 14.50 m.

When the sample size was 121, the minimum distance was 9.60 m and the maximum distance was 14.50 m.

When the sample sizes were 441 and 2601, the minimum distance was 9.60 m and the maximum distance was 14.50 m, respectively.

From the above verification results, it was found that the number of samples of the virtual antenna FA needs to be 121 or more in order to obtain the minimum delay time (minimum distance) and maximum delay time (maximum distance) that converge at a certain time. It was recognized before the simulation that the number of samples needs to be about 3000 to obtain the converged maximum delay time and minimum delay time, but in reality, an extremely small number of samples, less than 1/10 of that, is sufficient. By sufficiently reducing the number of samples, it becomes possible to perform the simulation at a higher speed. In order to satisfy both the reduction of the number of samples and stability (certainty), it is preferable to set the number of samples to 200 to 500, and more preferably to 300 to 400.

Incidentally, an experiment was conducted using an actual on-board radar to acquire the minimum and maximum distances within an actual predetermined range set similarly to the predetermined range P, with the number of samples being 3000. As a result of this experiment, the acquired minimum distance was 9.60 m and the maximum distance was 14.50 m, and it was confirmed that these matched the results of the simulation when the number of samples was 121 or more.

The probabilistic model is created for each combination of various parameters, such as the device positional relationship of the virtual vehicle with respect to the virtual antenna 1, more specifically, the azimuth of the virtual vehicle FV (e.g., changed in increments of 5 to 10 degrees), the direction of the virtual vehicle FV (e.g., changed in increments of 5 to 10 degrees for a full circumference of 360 degrees or a half circumference of 180 degrees), and the distance between the virtual antenna FA and the virtual vehicle FV (e.g., changed in increments of 0.1 to 0.5 m), as well as the shape of the virtual vehicle FV (differences in shape including size, e.g., differences between passenger cars, SUVs, trucks, buses, etc.). In the simulation, by programming in advance, it is possible to automatically and sequentially obtain probabilistic models corresponding to the combinations of the above parameters.

The number of probability models to be created is N1×N2×N3×N4, where N1 is the number of azimuth changes, N2 is the number of direction changes, N3 is the number of distance changes, and N4 is the number of shape changes. If no probability model matches the "certain relative relationship" between the virtual antenna FA and the virtual vehicle FV, a probability model for the "certain relative relationship" may be set by interpolation using multiple probability models that approximate the above-mentioned "certain relative relationship."

When evaluating the characteristics of an actual on-board radar before it is installed in a vehicle and used, the evaluation can be performed only by calculations using a pre-acquired probability model (no actual experiments are required). In other words, it is only necessary to substitute the lower limit (minimum delay time) and upper limit (maximum delay time) of the probability model shown in FIG. 4 into the range finder equation of the actual on-board radar to be verified. The minimum and maximum distances obtained by calculation are confirmed to evaluate the characteristics of the on-board radar. In this way, the technique of substituting the delay time allows the evaluation of the characteristics of the actual on-board radar to be performed in an extremely short time.

The delay time in the probabilistic model selected for verification is not limited to the lower and upper limits. For example, by selecting the delay time with the largest number of acquisitions , it is possible to know the most likely distance detected by the vehicle-mounted radar.

In addition, by substituting all the delay times in the probabilistic model into the distance calculation formula for the onboard radar and outputting the calculated distance obtained along with the number of delay times obtained in the probabilistic model , it is also possible to know the tendency of the detection distance to appear.

Since the delay time for each detected elementary wave is common regardless of the type or internal model of the onboard radar, such as pulse radar or FMCW radar (frequency modulated continuous wave radar), the probabilistic model can be commonly used regardless of the type or internal model of the onboard radar, making it highly versatile.

Next, a procedure for creating a probabilistic model by a computer that performs the above-mentioned simulation will be described with reference to the flowchart in Fig. 5. The flowchart in Fig. 5 corresponds to one simulation condition in which the relative positional relationship between the virtual antenna FA and the virtual vehicle FV shown in Fig. 4 is set to a certain positional relationship, and the virtual vehicle FV is set to a certain shape (for example, a four-door sedan type passenger car). Also, the change order of the virtual antenna FA within the predetermined range P is preset (for example, the first one is set to the origin position). In the following explanation, Q indicates a step.

First, in Q1, the position of the first virtual antenna FA is set. Then, in Q2, a simulation is performed. That is, detection element waves are irradiated (transmitted) one by one from the virtual antenna FA toward the virtual vehicle FV, and the delay time until the irradiated detection element waves are received by the virtual antenna FA is acquired and stored.

When the scanning of the virtual vehicle FV by the detected wave elements is completed, the process proceeds to Q3, where the delay times and the number of all the detected wave elements acquired are stored.

After Q3, in Q4, it is determined whether or not all position selections within the predetermined range P of the virtual antenna FA have been completed. Initially, the determination is NO, and the process proceeds to Q5. In Q5, the next position of the virtual antenna FA is selected (the position is changed). After this Q5, the process proceeds to Q2.

When the position selection of the virtual antenna FA is performed for all positions (in the embodiment, the number of positions is 121 or more, but close to 121), the determination in Q4 is YES. If the determination in Q4 is YES, in Q6, a probability model as shown in FIG. 4 is created based on the acquired delay times and their number. After that, in Q7, the probability model created in Q6 and the simulation conditions performed this time are associated and stored. The stored simulation conditions include the azimuth and direction of the virtual vehicle FV, the distance from the virtual antenna FA at the origin position, and the shape of the virtual vehicle, and may also include the number of samples of the position change of the virtual antenna FA and the change location, and the set predetermined range P.

The control shown in FIG. 5 is carried out many times while changing the simulation conditions.

6 shows an example of control for verifying the characteristic evaluation of an actual on-board radar using a probabilistic model. First, in Q11, an on-board radar to be verified is selected. The process of Q11 selects a distance calculation formula based on the delay time of the selected on-board radar.

After Q11, in Q12, a probability model corresponding to a condition to be verified is selected. Since there are usually multiple conditions to be verified, multiple probability models are selected. The selected multiple probability models are ranked.

After Q12, in Q13, a first probability model is selected from the selected plurality of probability models. Then, in Q14, the lower limit value of the delay time stored in the selected probability model is substituted into the distance calculation formula of the vehicle-mounted radar to calculate the lower limit distance (minimum distance). Next, in Q15, the upper limit value of the delay time stored in the selected probability model is substituted into the distance calculation formula of the vehicle-mounted radar to calculate the upper limit distance (maximum distance).

After Q15, in Q16, the calculated lower limit distance and upper limit distance are stored in correspondence with the currently selected probability model (simulation conditions of the model). After this, in Q17, the contents stored in Q16 are output (displayed on screen and/or printed out).
After Q17, in Q18, it is determined whether or not all probability models have been selected. Initially, the determination in Q18 is NO, and the process moves to Q19. In Q19, the next probability model is selected, and then the process moves to Q14.

When the processes in Q14 to Q17 have been carried out for all of the plurality of probabilistic models selected in Q12, the answer to Q18 becomes YES and the control ends.

Fig. 8 shows an example of a control system that also performs response control when an abnormality occurs in the distance detection by the vehicle-mounted radar S2. In Fig. 8, U is a controller (control unit) configured using a microcomputer. In addition to the above-mentioned vehicle-mounted camera S2 and vehicle-mounted radar S2, signals are input to this controller U from the headlamp switch S3 and a follow-up constant speed control switch (ON, OFF switch). The controller U also performs output control (vehicle speed control) of the engine S11, control of the brake S12 (automatic brake control and vehicle speed control), and control of the headlight S13 (high/low switching control). Furthermore, the controller U stores the above-mentioned numerous probability models as a database DB (stored in ROM).

Of the controls performed by the controller U, a control example related to distance detection by the on-board radar S2 mounted on the host vehicle V1 will be described with reference to the flowcharts in FIG.

First, in Q21 in Fig. 9, the vehicle-mounted camera S2 captures an image of the forward environment and recognizes a forward object (corresponding to the forward vehicle V2 in Fig. 1). Next, in Q22, the vehicle-mounted radar S2 detects the distance to the recognized forward object.

After Q22, in Q23, it is determined whether or not an abnormality has occurred in the distance detection. Abnormal states of distance detection include when distance detection itself is impossible, when the detected distance is abnormally large (the currently detected distance is abnormally large compared to the previously detected distance), or when the detected distance is abnormally small (the currently detected distance is abnormally large compared to the previously detected distance).

When the answer to the question Q23 is NO, that is, when there is no abnormality in the distance detection, the distance detected in Q22 is output in Q24. Based on the distance output in Q24, the constant speed following control, automatic brake control, and high/low headlight switching control are performed.

When the determination in Q23 is YES, in Q25, one probability model according to the current situation, i.e., the relative positional relationship between the vehicle V1 and the forward vehicle V2 and the shape of the forward vehicle V2, is selected as a specific probability model from among the multiple probability models. The specific probability model to be selected can be selected based on an image captured by the vehicle-mounted camera S1 of the situation immediately before an abnormality is detected in the distance detection by the vehicle-mounted radar S2. Note that, from the viewpoint of reducing the storage load and reducing costs in the controller U, multiple probability models can be stored in an information center (not shown), and the specific probability model can be acquired by communication with the information center as necessary.

After Q25, distance correction (distance determination) is performed in Q26 based on the selected specific probability model, as described below. After this, in Q27, the distance corrected in Q26 is output. Based on this distance output in Q27, adaptive cruise control, automatic brake control, and high/low headlight switching control are performed.

FIG. 7 is a diagram showing an example of distance correction in Q26. In FIG. 7, K1 is the delay time of the detection wave detected by the onboard radar S2. The distance corresponding to the detected delay time K1 is abnormally large and is larger than the upper limit (maximum delay time) of the delay time stored in the specific probability model. In this case, the distance corrected (determined) in Q26 can be determined as the distance corresponding to the delay time close to (including) the upper limit of the delay times stored in the specific probability model depending on the situation.

In Fig. 7, K2 is the delay time of the detection wave detected by the onboard radar S2. The distance corresponding to the detected delay time K2 is abnormally small and is smaller than the lower limit (minimum delay time) of the delay time stored in the specific probability model. In this case, the distance corrected (determined) in Q26 can be determined as the distance corresponding to the delay time close to (including) the lower limit of the delay times stored in the specific probability model depending on the situation.

Furthermore, when the vehicle radar S2 cannot detect the distance, in Q26, the delay time is appropriately set as the delay time stored in a specific probability model. More specifically, the delay time (or the distance corresponding to the delay time) to be set is preferably determined as the distance corresponding to the delay time around the time when the number of acquired items is the largest (hereinafter, also referred to as the peak distance). In particular, the peak distance can be set as the distance corresponding to the average delay time by averaging a plurality of delay times including the delay time when the number of acquired items is the largest and the delay times around the delay time.

Next, a preferred example of on-board control according to the distance to the forward vehicle V2 will be described, in which the detected distance is abnormal.

Fig. 10 shows an example of distance correction in Q26 in Fig. 9 when the following constant speed cruise control is performed. The following constant speed cruise control is performed when a sufficient distance is secured between the vehicle and the preceding vehicle V2, but unnecessary deceleration weakens the following degree and makes the driver feel uncomfortable. For this reason, the control is set in a direction that prevents the distance from becoming too small.

First, in Q31, it is determined whether or not distance detection is impossible. If the determination in Q31 is NO, then in Q32, it is determined whether or not the abnormal detected distance is smaller than the distance corresponding to the lower limit of the delay time stored in a specific probability model. If the determination in Q32 is YES, then in Q33, the distance is corrected (determined) to a distance corresponding to a delay time close to the lower limit in the specific probability model. In other words, by correcting the distance to a distance larger than the detected distance, it is possible to prevent unnecessary deceleration.

If the answer to the question Q32 is NO, the correction in Q33 is not performed, and the detected distance is used as it is for constant speed cruise control.

If the answer to the question Q31 is YES, then in Q34, the distance is corrected to the peak distance described above corresponding to the delay time with the greatest number of occurrences in a particular probability model and delay times in the vicinity thereof.

If the answer to Q32 is NO, it is assumed that the detected distance is greater than the distance corresponding to the upper limit in a particular probability model. In this case, the distance can be corrected to a distance corresponding to a delay time close to the upper limit in the particular probability model.

Fig. 11 shows an example of distance correction in Q26 of Fig. 9 when automatic braking control is performed. In this control example, distance correction is performed so that automatic braking is not started unnecessarily, while prioritizing collision avoidance during automatic braking.

First, in Q41, it is determined whether distance detection is impossible. If the result of the determination in Q41 is YES, in Q42, the peak distance in a specific probability model is set (this corresponds to Q34 in FIG. 10, and therefore a duplicated explanation is omitted).

If the determination in Q41 is NO, then in Q43 it is determined whether or not the automatic brake is operating. If the determination in Q43 is NO, that is, if the automatic brake control has not started, then the process proceeds to Q44. In Q44 it is determined whether or not the detected distance is smaller than the distance corresponding to the lower limit value in a specific probability model. If the determination in Q44 is YES, then in Q45 the distance is corrected to a distance corresponding to a delay time close to the lower limit value in the specific probability model. The process in Q45 corrects the distance to a larger distance, and the start of the automatic brake control is suppressed. If the determination in Q44 is NO, then the process is terminated (the detected distance is used as is).

When the determination in Q43 is YES, automatic brake control has been started. In this case, the process proceeds to Q46, where it is determined whether the detected distance is greater than the distance corresponding to the upper limit value of a specific probability model. When the determination in Q46 is YES, the distance is corrected in Q47 to a distance corresponding to a delay time close to the upper limit value in the specific probability model. Through the process of Q47, the distance is corrected to a smaller distance, which is a preferable distance setting for collision avoidance. When the determination in Q46 is NO, the detected distance is terminated as it is (the detected distance is used as it is).

11, it is possible not to execute control in a direction in which the correction distance becomes larger than the detected distance (control Q45, which prioritizes safety and makes the correction distance larger than the detected distance, is not executed). Also, when the answer to Q43 is YES, it is possible not to execute distance correction (for example, to continue automatic brake control that does not depend on the distance).

Fig. 12 shows an example of distance correction in Q26 of Fig. 9 when performing high/low headlight control. In this control example, distance correction is performed to prevent or suppress a situation in which the headlights are inadvertently set to the high state.

First, in Q51, it is determined whether distance detection is impossible. If the answer is YES in Q51, the peak distance in a specific probability model is set in Q52 (this corresponds to Q34 in FIG. 10, so a duplicated explanation is omitted).

If the answer to Q51 is NO, then in Q53 it is determined whether the detected distance is greater than the upper limit of a particular probability model. If the answer to Q46 is YES, then in Q54 the detected distance is corrected to a delay time that is closer to the upper limit of the particular probability model. By correcting the distance to a smaller distance, the headlights are prevented or suppressed from being turned on (preventing or suppressing a situation in which dazzling the occupants of the vehicle ahead is caused).

If the answer to the question Q53 is NO, the process ends (the detected distance is used as is for controlling the high/low switching).

Although the embodiment has been described above, the present invention is not limited to the embodiment, and appropriate modifications are possible within the scope of the claims. For example, the direction from the antenna position where the delay time is detected (the imaging direction of the vehicle-mounted camera S1) is not limited to the front, but can be an appropriate direction such as rearward, sideward, etc. The predetermined range P may include the up-down direction in addition to the forward/backward and left-right directions (corresponding to the up-down movement of the vehicle equipped with the vehicle-mounted radar, in this case, the predetermined range P is set in a three-dimensional box shape in the forward/backward, left-right, and up-down directions). The predetermined range P may be any one range or any two combination ranges of the forward/backward, left-right, and up-down directions. The delay time shown in the probability model can be excluded from the probability model if the number of acquired delay times is smaller than a predetermined value set in advance. The vehicle-mounted control device that performs control based on the detection distance is not limited to the case shown in the embodiment, and includes appropriate ones.

The sampling period for distance detection by the on-board radar is much smaller (shorter) than the control sampling period of the on-board device that performs control based on the detected distance. Therefore, distance correction based on a specific probability model may be performed on the condition that the detected distance is abnormal or undetectable a predetermined number of times or more within the control sampling period of the on-board device. Of course, the object of the present invention is not limited to what has been clearly stated, but also implicitly includes providing what is substantially preferable or expressed as an advantage.

The present invention is particularly useful when applied to vehicles that are controlled based on distance.

V1: Own vehicle V2: Front vehicle FA: Virtual antenna FV: Virtual vehicle (virtual reflecting object)
P: Predetermined range β1: Detected wave β2: Detected wave

Claims (7)

An onboard camera that captures images of the surrounding environment and recognizes target vehicles;
an on-vehicle radar for detecting a distance to the target vehicle;
a plurality of probability models that are created in advance by simulation in accordance with a plurality of delay times, which are the round trip times of a plurality of detection element waves between a virtual vehicle and a virtual antenna whose position is changed over a plurality of locations within a predetermined range, and that correspond to and store the number of acquired delay times for each different delay time, and that are created in accordance with a plurality of relative positional relationships between the virtual antenna and the virtual vehicle and differences in the shape of the virtual vehicle;
an abnormality detection means for detecting an abnormality in distance detection by the on-vehicle radar;
a selection means for selecting, when an abnormality is detected by the abnormality detection means, from among the multiple probability models, a specific probability model corresponding to a relative positional relationship with the target vehicle and a shape of the target vehicle acquired by the vehicle-mounted camera;
a correction means for correcting a distance to the target vehicle based on a delay time stored in the specific probability model when an abnormality is detected by the abnormality detection means;
A vehicle control device comprising: In claim 1,
A vehicle control device characterized in that, when the abnormality detection means detects an abnormal state in which distance detection is impossible using the onboard radar, the correction means determines the distance corresponding to a delay time selected from the delay times stored in the specific probability model as the corrected distance. In claim 2,
A vehicle control device characterized in that the correction means determines the distance after correction to be the distance corresponding to the delay time closest to the delay time stored in the specific probabilistic model that is most frequently obtained . In any one of claims 1 to 3,
A vehicle control device characterized in that, when the abnormality detection means detects an abnormal state in which the detection distance by the onboard radar is abnormally large, the correction means determines the distance after correction to be the upper limit value of the longest delay time among the delay times stored in the specific probability model or a distance corresponding to the vicinity of the upper limit value. In any one of claims 1 to 4,
A vehicle control device characterized in that, when the abnormality detection means detects an abnormal state in which the detection distance by the onboard radar is abnormally small, the correction means determines the distance after correction to be the lower limit value of the delay time with the smallest delay time among the delay times stored in the specific probability model, or the distance corresponding to the vicinity of the lower limit value. In any one of claims 1 to 5,
a vehicle control device, characterized in that the specified range is any one of a range in the forward/backward direction, a left/right direction, or an up/down direction of the virtual antenna, a range consisting of any two of these, or a range consisting of all combinations. In any one of claims 1 to 6,
a vehicle control device, characterized in that the multiple relative positional relationships set when creating the multiple probabilistic models are a combination of an orientation of the virtual vehicle with respect to the virtual antenna, a direction of the virtual vehicle, and a distance from the virtual antenna to the virtual vehicle.

Images