JP7294010B2 - Driving support device - Google Patents

Description

The present disclosure relates to driving assistance devices.

In Patent Document 1, the presence or absence of the driver's intention to drive is determined based on the frequency of occurrence of saccades, which are minute eye movements that differ from the visual behavior of the driver. is described. In this driving support device, it is determined that there is no driving intention when the frequency of occurrence of saccades is high.

JP 2017-187825 A

However, as a result of detailed studies by the inventors, the following problems were found. That is, with the device described in Patent Document 1, it is difficult to accurately determine whether or not the driver intends to drive when the visual action is being performed for a purpose other than the driving operation. For example, when the driver intentionally performs a visual action other than the driving operation, such as looking aside at the navigation screen, the occurrence frequency of saccades is considered to be low. Therefore, there is a possibility that it may be determined that there is a driving intention even though there is actually no driving intention.

One aspect of the present disclosure provides a driving assistance device that can more accurately determine whether or not there is a driving intention.

The present invention proposes a new concept of future predicted time, and provides a device that captures the presence or absence of the driver's driving intention. The future prediction time is an index indicating how far ahead the driver predicts the future while driving. When a driver drives with intention, he/she drives by predicting the future several seconds ahead. For example, when changing lanes, the driver sets a target position to reach after changing lanes, predicts the vehicle trajectory to reach the target, and performs a steering operation. In addition, when the vehicle is to be stopped, the driver determines a target stop position, predicts the vehicle behavior, and operates the brakes in order to stop at the target position. At this time, the time required from setting the target to reaching the target corresponds to the future prediction time. Even if there is no explicit target, the concept of future prediction time applies. For example, when driving with a constant distance to the vehicle in front, the driver must anticipate changes in the situation several seconds ahead in order to respond to ever-changing conditions (such as the speed of the vehicle ahead and road gradient). while performing accelerator operation or brake operation. In this way, regardless of the presence or absence of an explicit target, the driver performs operations while predicting the next few seconds. And vehicle behavior changes with regularity. Due to the regularity, it is possible to predict the driving behavior of the driver, and the future prediction time can be obtained from the prediction result.

One aspect of the present disclosure is a driving support device mounted on a vehicle, which includes a future prediction time calculation unit (S101 to S106) and a driving intention presence/absence determination unit (S107, S108). The future prediction time calculation unit is configured to calculate a future prediction time based on vehicle state information representing at least one of the driving operation of the driver and the behavior of the vehicle. The driving intention presence/absence determination unit is configured to determine whether or not to perform at least one of notification to the driver and driving assistance based on the future predicted time.

According to such a configuration, it is possible to more accurately determine the presence or absence of the driving intention.

1 is a block diagram showing the configuration of a driving support device; FIG. (A) is a diagram for explaining a lane change trajectory predicted by a driver. (B) is a diagram for explaining the future prediction time in a scene of lane change. FIG. 4 is a diagram for explaining the time series of the first period and the second period; 4 is a flowchart of driving intention determination processing; 7 is a graph showing the ratio of the number of experimental trials in which a contact accident with an object to be avoided occurred for each future prediction time; FIG. 10 is a graph showing the ratio of experimental trial counts in which driving maneuvers for avoidance behavior were started immediately after the appearance of an object to be avoided, for each future prediction time; FIG.

Exemplary embodiments of the present disclosure are described below with reference to the drawings.
[1. composition]
A driving assistance device 1 shown in FIG. 1 is installed in a vehicle such as an automobile, and is a device for assisting driving according to the presence or absence of a driving intention of a driver who drives the vehicle. Hereinafter, the vehicle in which the driving assistance device 1 is mounted is referred to as the own vehicle.

In addition to the driving support device 1 , the own vehicle is equipped with a vehicle sensor 21 , a surroundings monitoring sensor 22 , a notification unit 23 , and an operation intervention unit 24 .
The vehicle sensor 21 is for detecting vehicle information such as driving operation information for the own vehicle and information regarding the running state of the own vehicle. In this embodiment, as the vehicle sensors 21, a steering angle sensor 21a, an accelerator opening sensor 21b, a brake pedal operation sensor 21c, and a vehicle speed sensor 21d are mounted. The steering angle sensor 21a is a sensor for detecting the steering angle of the steering of the own vehicle. The accelerator opening sensor 21b is a sensor for detecting the depression amount of the accelerator pedal. The brake pedal operation sensor 21c is a sensor for detecting the on/off state of the brake. The vehicle speed sensor 21d is a sensor for detecting the travel speed of the own vehicle.

The surroundings monitoring sensor 22 is for detecting surrounding information such as information about the surrounding environment of the own vehicle. In this embodiment, as the surroundings monitoring sensor 22, a lateral position detector 22a and an inter-vehicle distance detector 22b are mounted. The lateral position detection unit 22a is for detecting how much the vehicle is traveling to the left or right from the center of the lane. In this embodiment, a front camera is mounted as the lateral position detector 22a. The front camera captures an image of a predetermined range ahead of the vehicle. An image captured by the front camera is used for processing for detecting road markings such as lane markings. The inter-vehicle distance detection unit 22b is for detecting the inter-vehicle distance from the preceding vehicle. In this embodiment, a millimeter wave radar is installed as the following distance detector 22b. A millimeter wave radar transmits millimeter waves to the surroundings of the own vehicle and receives reflected waves reflected by an object. Objects include vehicles other than the host vehicle, such as preceding vehicles. A survey wave sensor such as sonar or LIDAR may be used as the inter-vehicle distance detection unit 22b. Note that LIDAR is an abbreviation for Light Detection and Ranging. Hereinafter, the vehicle information detected by the vehicle sensor 21 and the peripheral information detected by the peripheral monitoring sensor 22 are collectively referred to as vehicle state information. The vehicle sensor 21 and the surroundings monitoring sensor 22 are configured to be able to output vehicle state information to the driving assistance device 1 directly or through an in-vehicle network such as CAN (registered trademark).

The driving support device 1 is configured mainly by a well-known microcomputer having a CPU, ROM, RAM, flash memory, etc. (not shown). The CPU executes a program stored in a ROM, which is a non-transitional substantive recording medium. A method corresponding to the program is executed by executing the program. Specifically, the driving assistance device 1 executes driving intention determination processing shown in FIG. 4, which will be described later, according to the program. Note that the driving assistance device 1 may include one microcomputer, or may include a plurality of microcomputers.

The driving support device 1 includes a future prediction time calculation unit 11 and a driving intention presence/absence determination unit 12 as functional blocks realized by the CPU executing a program, that is, virtual components. The method of realizing the functions of the units included in the driving assistance device 1 is not limited to software, and a part or all of the functions may be realized using one or more pieces of hardware. For example, when the above functions are realized by an electronic circuit that is hardware, the electronic circuit may be realized by a digital circuit, an analog circuit, or a combination thereof.

The future prediction time calculator 11 calculates the future prediction time. For example, in the scene of changing lanes, as described above, the time required from setting a target to reaching the target corresponds to the future prediction time. In a lane change scene as shown in (A) of FIG. 2, as shown in (B) of FIG. 2, the predicted future time is represented by L[m]/V[m/s]. The future prediction time calculated by the future prediction time calculation unit 11 is used by the driving intention presence/absence determination unit 12 when determining whether or not there is a driving intention. The future prediction time calculation unit 11 includes a similarity calculation unit 111 and an estimation unit 112 .

The similarity calculation unit 111 sets a plurality of periods in the past with the current time as a reference and having different lengths in stages. Next, the similarity calculation unit 111 divides each of the plurality of set periods into a first period that is a period before a predetermined reference point in time, a second period that is a period after the reference point in time, and split into In this embodiment, a common reference point in time is used for each of a plurality of periods. That is, the second period is a period immediately preceding the current time and has a constant length, while the first period is a period before the second period and has a different length in stages. Next, for each of the plurality of first periods, the similarity calculation unit 111 calculates prediction information that predicts the vehicle state information in the second period based on the vehicle state information in the first period. Then, the similarity calculation unit 111 calculates the similarity between each of the calculated pieces of prediction information and the measured information, which is the vehicle state information actually measured in the second period.

As a method of calculating prediction information, for example, N-gram model technology used in speech recognition or the like can be used. That is, in a situation where the driver is operating the vehicle with the intention of driving, the vehicle state information changes with regularity according to the type of driving scene such as starting or changing lanes. Therefore, by statistically processing time-series changes in vehicle state information using the N-gram model technique, it is possible to obtain the probability of transition from one vehicle state information to the next vehicle state information. For example, in a starting scene, which is a driving scene when starting a vehicle, the driver starts depressing the accelerator pedal, further depresses the accelerator pedal, keeps the accelerator pedal depression amount constant for a while, and then slightly releases the accelerator pedal. Do the driving operation. At this time, the accelerator opening degree in the vehicle state information has a regularity that it gradually increases from 0%, and after holding the same opening degree for a certain period of time, it decreases. Therefore, it is possible to obtain the probability of transition from a certain accelerator opening state to the next accelerator opening state by statistically processing the extent to which it applies to the starting scene based on the time-series changes in the vehicle state information. It becomes possible. In this embodiment, prediction information is represented by a plurality of types of vehicle state information in the second period predicted based on the vehicle state information in the first period and the occurrence probability of transition to each vehicle state information. . Taking the accelerator opening as an example, the occurrence probability that the accelerator opening gradually increases from 0%, the occurrence probability that the accelerator opening is maintained at the same opening for a certain period of time, and the accelerator opening The probability of occurrence of transition to a smaller state is calculated as prediction information. Further, in the present embodiment, the probability of occurrence of transition from vehicle state information to actual measurement information in the first period is calculated as the degree of similarity. Taking the accelerator opening as an example, if the measured information indicates that the accelerator opening gradually increases from 0%, among the plurality of occurrence probabilities calculated as the prediction information, the accelerator opening gradually increases from 0%. The probability of occurrence of transition to a larger state is calculated as the degree of similarity.

The estimation unit 112 estimates future prediction time based on the similarity calculated by the similarity calculation unit 111 . The estimation unit 112 estimates the length of the first period corresponding to the prediction information calculated to have the highest similarity among the plurality of pieces of prediction information as the future prediction time.

An example of processing in the similarity calculation unit 111 and the estimation unit 112 will now be described with reference to FIG. In the example of FIG. 3, three pieces of prediction information are calculated for the sake of simplicity of explanation. That is, as past periods based on the current time, a period T20 from the first time t2 to the current time t0, a period T30 from the first time t3 to the current time t0, and a period T30 from the first time t4 to the current time t0. A period T40 is set. In this example, time t4, time t3, time t2, time t1, and time t0 are set at regular intervals in time series in this order.

First, the similarity calculation unit 111 calculates a predetermined period T20 from a first time point t2 to a second time point t1, a first time period 302a from a first time point t2 to a second time point t1, and a predetermined period T20 from a second time point t1 to the current time point. and a second period 302b until t0. Similarly, for the predetermined period T30, the similarity calculation unit 111 calculates a first period 303a from the first time t3 to the second time t1 with the second time t1 as a reference time and and a second period 303b up to the current time t0. Similarly, for the predetermined period T40, the similarity calculation unit 111 uses the second time point t1 as a reference time point, the first period 304a from the first time point t4 to the second time point t1, and the period from the second time point t1 to and a second time period 304b up to the current time t0.

Next, the similarity calculation unit 111 calculates prediction information A, which is vehicle state information in the second period 302b predicted based on the vehicle state information in the first period 302a. Similarly, the similarity calculation unit 111 calculates the prediction information B, which is the vehicle state information in the second period 303b predicted based on the vehicle state information in the first period 303a, and the vehicle state in the first period 304a. Prediction information C, which is vehicle state information in the second period 304b predicted based on the information, is calculated.

Here, it is assumed that the occurrence probability, which is actually measured information, is calculated as 0.2 for prediction information A, 0.5 for prediction information B, and 0.7 for prediction information C. FIG. In this case, the estimating unit 112 estimates the length of the first period 304a corresponding to the predicted information C calculated to have the highest occurrence probability as the actually measured information, as the future predicted time.

The driving intention presence/absence determination unit 12 determines whether or not to notify the driver and to perform driving assistance. In performing this determination, the driving intention presence/absence determination unit 12 determines whether or not the driver has a driving intention based on the future prediction time calculated by the future prediction time calculation unit 11 . A state in which the driver has a driving intention means a state in which the driver can concentrate on the driving operation. The state in which the driver does not intend to drive means a state in which the driver cannot concentrate on the driving operation, such as when the driver feels sleepy or looks aside. The driving intention presence/absence determination unit 12 determines whether or not there is a driving intention by determining whether or not the predicted future time falls within a predetermined range. This is based on the relationship between the driving intention and the predicted future time found by the present inventor.

The present inventor investigated the relationship between driving intention and future prediction time through driving simulator experiments. Based on the results, based on the concept described later, the range of the future prediction time when the driver has the driving intention was determined. The predetermined range will be described with reference to FIG. 5 and FIG. 6, which are experimental results. In this experiment, the object to be avoided was caused to appear at several stages of timing based on TTC (Time-To-Collision: the value obtained by dividing the distance between the vehicle and the object to be avoided by the relative speed) while driving. It is assumed that the appearance timing of each trial corresponds to the future prediction time. For example, when the object to be avoided appears at TTC=2.0 seconds, the avoidance action is immediately started, and if the avoidance is successful, this corresponds to action with a future prediction time of 2.0 seconds. FIG. 5 is a graph showing the ratio of the number of experimental trials in which a contact accident with an object to be avoided occurred by future prediction time. The accident occurred because the driver's future prediction time was longer than the time until the vehicle collided with the object to be avoided, and accidents occurred in almost all trials with a future prediction time of 1 second. Therefore, in this embodiment, the future prediction time of 1.5 seconds is set as the lower limit. Moreover, FIG. 6 is a graph showing the ratio of the number of experimental trials in which a driving operation for an avoidance action was started immediately after the appearance of the object to be avoided, for each future prediction time. Considering implementation in the system, we set an upper limit from the viewpoint of redundancy in future prediction. As shown in FIG. 6, in a region where the future prediction time is 4.5 seconds or more (consideration of errors such as oversight), driving behavior is not immediately started in some trials. This means that the driver did not perform an avoidance operation as the most recent driving action because there was time left until the avoidance, and this indicates that the future prediction time is redundant. For this reason, 4.5 seconds is set as the upper limit in this embodiment.

Based on such knowledge, the driving assistance device 1 of the present embodiment determines that the driver has an intention to drive when the future prediction time is within a predetermined range, and does not notify the driver or provide driving assistance. I judge. Further, when the future prediction time is shorter than the lower limit value or longer than the upper limit value of the predetermined range, the driving support device 1 determines that the driver has no driving intention, and notifies the driver and provides driving support. Decide to do it.

The notification unit 23 is configured to notify the driver based on the determination by the driving intention presence/absence determination unit 12 . For example, the driver's attention is called using voice or screen display. The notification unit 23 performs appropriate notification according to the driving scene.

The operation intervention unit 24 is configured to perform driving assistance based on the determination by the driving intention presence/absence determination unit 12 . For example, the driving operation is assisted using the ACC function, the LKA function, and the like. The operation intervention unit 24 executes appropriate driving assistance according to the driving scene. ACC is an abbreviation for Adaptive Cruise Control. LKA is an abbreviation for Lane Keeping Assist.

[2. process]
The driving intention determination processing executed by the driving support device 1 will be described using the flowchart of FIG. 4 . The driving intention determination process is repeatedly executed at a predetermined cycle after the ignition switch of the host vehicle is turned on.

First, in S101, the driving support device 1 sets the n-th past period based on the current time. Note that the initial value of n is 1 in this embodiment. In this embodiment, the driving support device 1 sets the period from 0.5×(n+1) seconds ago to the current time as the n-th past period. For example, when n=3, the driving support device 1 sets the period from 2 seconds ago to the present time as the third past period.

Subsequently, in S102, the driving support device 1 divides the period set in S101 into the first period and the second period as described above. In this embodiment, the reference point in time for dividing the first period and the second period is assumed to be 0.5 seconds before the current point in time. For example, when n=3, the driving assistance device 1 defines the third past period as the first period from 2 seconds ago to 0.5 seconds ago and the period from 0.5 seconds ago to the present time. and a second period that is a period of

Subsequently, in S103, the driving support system 1 calculates prediction information, which is vehicle state information for the second period predicted based on the vehicle state information for the first period.
Subsequently, in S104, the driving assistance device 1 determines whether or not the processes of S101 to S103 have been executed a predetermined number of times. In this embodiment, the driving support device 1 sets the predetermined number of times to 10.

If the driving support system 1 determines in S104 that the processes of S101 to S103 have not been executed a predetermined number of times, it adds 1 to n and returns to S101.
On the other hand, when it is determined in S104 that the processes of S101 to S103 have been performed a predetermined number of times, the driving support system 1 proceeds to S105 and calculates the degree of similarity between each of the prediction information and the actual measurement information. Note that S101 to S105 correspond to the processing of the similarity calculation unit 111. FIG.

Subsequently, in S106, the driving support device 1 compares the similarities, and estimates the first period corresponding to the prediction information with the highest similarity as the future prediction time. Note that S106 corresponds to processing by the estimation unit 112 .

Subsequently, in S107, the driving support device 1 determines whether or not the future prediction time falls within a predetermined range.
When the driving assistance device 1 determines in S107 that the future prediction time does not fall within the predetermined range, the driving assistance device 1 proceeds to S108, determines to perform notification and driving assistance, and then performs the driving intention determination processing of FIG. finish.

On the other hand, when the driving support device 1 determines in S107 that the future prediction time falls within the predetermined range, it ends the driving intention determination process of FIG. It should be noted that S107 and S108 correspond to the processing of the driving intention presence/absence determination unit 12 .

[3. effect]
According to this embodiment detailed above, the following effects are obtained.
(3a) The driving support device 1 calculates the predicted future time based on the vehicle state information, which is information representing at least one of the driving operation of the driver and the behavior of the vehicle. Then, the driving support device 1 determines whether or not there is a driving intention based on the predicted future time. On the other hand, when the presence or absence of driving intention is determined based on the frequency of occurrence of saccades, as described above, there is a possibility that it is impossible to distinguish between driving intention and other intentions. A similar problem may occur when determining the presence or absence of driving intention using a scale based on the physical condition of the driver other than the occurrence frequency of saccades. According to the configuration of this embodiment, there is a problem that it is not possible to distinguish between driving intention and other intentions when determining the presence or absence of driving intention using a scale based on the physical condition of the driver, such as the frequency of occurrence of saccades. is less likely to occur. Therefore, it is possible to more accurately determine the presence or absence of the driving intention compared to the case of determining the presence or absence of the driving intention using a scale based on the physical condition of the driver.

(3b) The driving support device 1 determines that the driver intends to drive when the predicted future time is within a predetermined range. This is based on the knowledge that when the driver has an intention to drive, the driver predicts the future within a predetermined range and drives the vehicle. It is also based on the knowledge that the time that the driver can predict is the same regardless of the driving scene. Therefore, there is no need to perform processing to determine whether or not there is a driving intention using different logic for each specific driving scene. Therefore, it is possible to determine the presence or absence of the driving intention in general compared to the case where the presence or absence of the driving intention is determined using different logic for each specific driving scene.

(3c) If the future prediction time is shorter than the lower limit value of the predetermined range, the driving support device 1 determines that the driver has no driving intention. According to such a configuration, it is possible to eliminate the case of an impulsive action that cannot be foreseen.

(3d) If the future prediction time is longer than the upper limit value of the predetermined range, the driving support device 1 determines that the driver has no driving intention. According to such a configuration, it is possible to eliminate the case where the degree of similarity between the predicted information and the actually measured information becomes high by chance even though the future exceeds human processing ability.

[4. Other embodiments]
Although the embodiments of the present disclosure have been described above, it is needless to say that the present disclosure is not limited to the above embodiments and can take various forms.

(4a) In the above embodiment, the operation intervention unit 24 executes the ACC function and the LKA function as driving support functions. However, for example, instead of the driving support function, the operation intervention unit 24 may be configured to execute the automatic driving function.

(4b) In the above embodiment, the estimation unit 112 estimates the first period corresponding to the prediction information calculated to have the highest degree of similarity as the future prediction time. However, for example, a period obtained by adding the first period and the second period corresponding to the prediction information calculated to have the highest similarity may be estimated as the future prediction time.

(4c) In the above embodiment, the similarity calculation unit 111 calculates a plurality of types of vehicle state information in the second period predicted based on the vehicle state information in the first period and occurrences of transition to each vehicle state information. We represent predictive information by probabilities. However, for example, the prediction information may be represented by a driving scene identified based on the vehicle state information in the second period predicted based on the vehicle state information in the first period. In this case, the similarity calculation unit 111 also expresses the measured information by driving scenes, and calculates the similarity by comparing the driving scenes. In addition, the similarity calculation unit 111 calculates a value based on a combination of conditions such as the steering angle of the vehicle's steering wheel, the accelerator opening of the vehicle's vehicle, and how much the vehicle's vehicle is traveling to the left or right from the center of the lane. Then, a process of converting the vehicle state information into a driving scene is performed.

(4d) In the above embodiment, the technique of the N-gram model is used as a method of calculating prediction information. However, other methods may be used to calculate the prediction information. For example, hidden Markov models can be used.

(4e) The function of one component in the above embodiments may be distributed as multiple components, or the functions of multiple components may be integrated into one component. Also, part of the configuration of the above embodiment may be omitted. Also, at least a part of the configuration of the above embodiment may be added, replaced, etc. with respect to the configuration of the other above embodiment.

1... Driving support device, 11... Future prediction time calculation unit, 12... Driving intention presence/absence determination unit.

Claims (4)

A driving support device mounted on a vehicle,
A future that is an index that indicates how far into the future the driver is driving based on vehicle state information that is information representing at least one of the driver's driving operation and the behavior of the vehicle. a future prediction time calculation unit (S101 to S106) configured to calculate a prediction time;
a driving intention presence/absence determining unit (S107, S108) configured to determine whether or not to perform at least one of notification to the driver and driving assistance based on the predicted future time;
with
The future prediction time calculation unit
A period from the second time point to the current time point predicted based on the vehicle state information in each of a plurality of first time periods, which is a period from a plurality of first time points in the past to a second time point prior to the current time It is configured to calculate a degree of similarity between each of a plurality of prediction information, which is the vehicle state information in a second period, and actual measurement information, which is the vehicle state information actually measured in the second period. a similarity calculation unit (S101 to S105);
An estimating unit configured to estimate the future prediction time according to the length of the first period corresponding to the prediction information for which the highest degree of similarity with the measured information is calculated among the plurality of prediction information. (S106), and a driving support device. The driving support device according to claim 1 ,
The similarity calculation unit identifies each driving scene, which is a type of time-series change with regularity of the vehicle state information identified based on each of the plurality of prediction information, and the actual measurement information. A driving assistance device that calculates a degree of similarity with the driving scene. The driving support device according to claim 1 or claim 2 ,
If the future prediction time is shorter than the lower limit value of the predetermined range or longer than the upper limit value of the predetermined range, the driving intention presence/absence determination unit performs at least one of notification to the driver and driving assistance. A driving support device that determines to perform. A driving support device mounted on a vehicle,
A future that is an index that indicates how far into the future the driver is driving based on vehicle state information that is information representing at least one of the driver's driving operation and the behavior of the vehicle. a future prediction time calculation unit (S101 to S106) configured to calculate a prediction time;
a driving intention presence/absence determining unit (S107, S108) configured to determine whether or not to perform at least one of notification to the driver and driving assistance based on the predicted future time;
with
If the future prediction time is shorter than the lower limit value of the predetermined range or longer than the upper limit value of the predetermined range, the driving intention presence/absence determination unit performs at least one of notification to the driver and driving assistance. A driving support device that determines to perform.

Images