JP4791874B2 - Driving support device and driving action determination device - Google Patents

Description

  The present invention relates to a driver model creation device and a driving support device, and relates to, for example, a driver model creation device that serves as an evaluation criterion for a driving state, and a device that performs driving state evaluation and driving support using a driver model.

Various proposals have been made on the modeling of the driving operation of a vehicle driver (driver) and its application.
For example, the technique described in Patent Document 1 proposes a technique for evaluating the risk of an intersection road by a driver model using fuzzy rules or a neural network.
JP 2002-140786

By using such a driver model, the driving state estimated from the driver model is estimated as a normal driving state, and the current driving state is compared with this to evaluate the current driving state. It can be performed.
However, the driving state estimated from the driver model is not always a normal driving state for the driver.
Even if a specific driver actually drives the vehicle and collects driving state data and creates a driver model in advance based on this data, driving in a normal state is not always performed. There wasn't.

Accordingly, a first object of the present invention is to create a driver model with higher accuracy that serves as an evaluation criterion for driving conditions.
It is a second object of the present invention to provide a driving support device that uses the driver model to perform more accurate driving state evaluation and driving support.
A third object is to evaluate the driving state with higher accuracy with respect to the driving behavior of the driver.

(1) In the first aspect of the invention, a driver model acquisition unit that acquires, as a driver model, a Gaussian mixture model obtained by calculating travel data including a plurality of types of feature values of a driving operation in a normal state by an EM (Expectation Maximization) algorithm ; Using the acquired driver model, the driving operation estimating means for estimating a driving operation that is normally operated in a normal state, the estimated driving operation, and the driving operation based on the current driving operation information, A plurality of characteristics obtained by the driver model acquisition means, the driving behavior estimation means comprising: driving behavior determination means for determining behavior; and driving assistance means for performing driving assistance according to the determined driving behavior. Measure travel data excluding specific features out of the quantity, and measure against this travel data By calculating the maximum posterior probability in the driver model, the second object is achieved by the driving support device that estimates the specific feature amount .
(2) In the invention according to claim 2, in the driving support device according to claim 1, the plurality of types of feature values of the driving operation acquired by the driver model acquiring means are an accelerator operating amount, a vehicle speed, a brake operating amount, a steering Travel data comprising a plurality of types of feature amounts such as an operation amount, an inter-vehicle distance, and an acceleration.
(3) In the third aspect of the present invention, in the driving support apparatus according to claim 1 or claim 2, wherein the driver model obtaining means, from the driver model of the driving operation in a normal state created for each traveling environment, A driver model corresponding to the current driving environment is acquired.
(4) In the invention according to claim 4 , in the driving support apparatus according to claim 1, claim 2 or claim 3 , further comprising a driver state determining means for determining the state of the driver from the biological information of the driver. The driving support means performs driving support according to the determined driving behavior and the determined driver state.
(5) In the invention according to claim 5 , in the driving support apparatus according to claim 1 , claim 2, claim 3 or claim 4 , the driving support means is caution by voice or image according to the determination contents. At least one of driving assistance among arousal, information provision, vibration, and rest area guidance is provided.
(6) In the invention described in claim 6, a driver model acquisition unit that acquires, as a driver model, a Gaussian mixture model obtained by calculating travel data including a plurality of types of feature values of a driving operation in a normal state by an EM (Expectation Maximization) algorithm ; Using the acquired driver model, the driving operation estimating means for estimating a driving operation that is normally operated in a normal state, the estimated driving operation, and the driving operation based on the current driving operation information, Driving behavior determining means for determining an action, wherein the driving operation estimating means measures driving data excluding a specific feature quantity among a plurality of types of feature quantities acquired by the driver model acquisition means, and By calculating the maximum posterior probability in the driver model for the driving data, The third object is achieved by a driving behavior determination device that estimates a specific feature amount .
(7) In the invention according to claim 7, in the driving behavior determination device according to claim 6, a plurality of types of feature values of the driving operation acquired by the driver model acquisition means are an accelerator operation amount, a vehicle speed, a brake operation amount, The present invention is characterized in that it is travel data composed of a plurality of types of feature amounts of steering operation amount, inter-vehicle distance, and acceleration.

In the present invention of claim 1 to claim 5, from the driving operation that is normally operated in the normal state estimated using the driver model of the driving operation in the normal state, and the driving operation based on the current driving operation information, Since the driving behavior of the driver is determined and driving assistance is performed according to the determined driving behavior, it is possible to evaluate the driving state with higher accuracy and perform driving assistance.
In the inventions according to claim 6 and claim 7 , from the driving operation that is normally operated in the normal state estimated using the driving model of the driving operation in the normal state, and the driving operation based on the current driving operation information, Since the driving behavior is determined, it is possible to evaluate the driving state with higher accuracy.

Hereinafter, preferred embodiments of the driver model creation device and the driving support device of the present invention will be described in detail with reference to FIGS. 1 to 23.
(1) Outline of Embodiment In this embodiment, it is recognized whether or not the driver is in a normal state by detecting the biological information of the driver. And while driving, the driver collects driving state data (own vehicle information such as accelerator, brake, steering operation amount, vehicle speed, inter-vehicle distance, acceleration, etc.). A driver model is created by extracting the driving part.
This makes it possible to automatically create a normal driver model without making the driver aware of it.
In addition, since the driver model is created only when the driver is driving in a normal state based on the driver's biological information as a normal driving behavior, a more accurate neutral driver model can be obtained.

  In this embodiment, for example, on a three-lane one-way national road, when the signal turns blue from the right turn exclusive lane, there is an oncoming vehicle, and there are pedestrians on the pedestrian crossing, A driver model is created for each scene (situation) in the surrounding environment of the vehicle.

Also, by comparing in real time the normal driving behavior estimated from the created driver model with the current driver's driving behavior, the current driver's driving behavior is “as usual” or deviates. To monitor.
As an index for comparing the “normal” driving with the current driving, for example, the “reaction speed” and “flicker” of the driver are used.

  Furthermore, in this embodiment, not only the change in driving behavior based on the driver model is evaluated, but also the change in biological information is taken into account, so that information indicating the driver's state is determined in a composite manner, and the driver is more accurately detected. Detects fatigue and reduced attention.

As a result, when there is a deviation from the driver's original driving behavior, a safe driving support suitable for the person can be provided by alerting, warning and presenting information.
In addition, it is possible to detect the driver state at a precursor stage before the obvious fatigue or a reduction in attention, and it is possible to provide advanced guidance such as prompting a break before fatigue reaches a peak.

  In the present embodiment, by using a GMM (Gaussian mixture model) as a driver model, a driver model for each driver can be easily generated, and further, driving operation behavior is calculated by calculation that maximizes a conditional probability. Easily estimate and output.

That is, the driver model creation device, the driving support device, and the driving behavior determination device according to the present embodiment include a plurality of types of feature amounts such as an accelerator operation amount, a vehicle speed, a brake operation amount, a steering operation amount, an inter-vehicle distance, and an acceleration. A Gaussian mixture model calculated by an EM (Expectation Maximization) algorithm using travel data as learning data is adopted as a driver model.
This Gaussian mixture model is composed of the parameters of the joint probability density function obtained by calculating the joint probability density distribution by the EM algorithm, and for each driver as necessary, for the driver's accelerator operation, for brake operation, It is generated for each estimated feature amount for the inter-vehicle distance maintenance range.

  Then, travel data Y (= y1, y2,...) Excluding a specific feature amount x among a plurality of feature amounts used in the driver model is measured, and a maximum posterior probability in the driver model for the travel data Y is calculated. Thus, the feature quantity x is estimated.

  For example, the driver model in the same situation as the driving environment (situation) around the vehicle is used, the current own vehicle state is input to the driver model, and the subsequent driving state (for example, feature amount x = accelerator operation amount) By estimating the change and comparing it with the actual driving state, it is determined whether or not there is an operation delay, an operation fluctuation, or the like.

(2) Details of Embodiment FIG. 1 shows a configuration of a driving support device to which a driver model creation device is applied.
The driving support device includes an ECU (electronic control unit) 10, a host vehicle information acquisition unit 11, a host vehicle surrounding environment information acquisition unit 12, a biological information acquisition unit 13, an information provision unit 14, a driver model processing unit 15, and a data storage unit 16. It has.
Note that the configuration of the driving support apparatus described with reference to FIG. 1 is not all necessary, but describes each unit and apparatus that can be used to create a driver model and perform driving support in this embodiment. It is possible to configure the driving support device by appropriately selecting according to the function of the driving support device to be adopted, and it is possible to additionally use other devices and devices having similar functions. is there.

The ECU 10 is composed of a computer system including CPU, ROM, RAM, and interface units.
The ECU 10 monitors the driver driving behavior based on the acquired information of the host vehicle information acquiring unit 11, monitors the driver biological information based on the acquired information of the biological information acquiring unit 13, and provides the driver assist content information providing unit 14 as driving assistance to the information providing unit 14. Instructions. The ECU 10 also supplies the driver model processing unit 15 with data necessary for creating and outputting the driver model.

The own vehicle information acquisition unit 11 includes a steering wheel steering angle sensor 111, an accelerator pedal sensor 112, a brake pedal sensor 113, a speedometer 114, an acceleration sensor 115, an electric operation state acquisition unit 116, a timer 117, and other sensors.
FIG. 2 exemplifies host vehicle information as driving operation information acquired by the host vehicle information acquiring unit 11.
As shown in FIG. 2, the steering angle sensor 111 detects a steering operation amount (angle), the accelerator pedal sensor 112 detects an accelerator operation amount, the brake pedal sensor 113 detects a brake operation amount, and the speedometer 114 detects a vehicle speed.
The acceleration sensor 115 detects yaw axis acceleration, pitch axis acceleration, and roll axis acceleration.

The electric operation status acquisition unit 116 detects a winker operation status, a light operation status, and a wiper operation status.
The timer 117 measures various times such as operation time and operation time.

The own vehicle surrounding environment information acquisition unit 12 includes a vehicle surrounding information acquisition unit 121, a road information acquisition unit 122, and a network unit 123.
The vehicle surrounding information acquisition unit 121 includes various sensors such as an infrared sensor, a millimeter wave sensor, an ultrasonic sensor, an image recognition device, and an inter-vehicle distance sensor. The image recognition device performs image processing of an outside image captured by the image input device, and recognizes obstacles around the vehicle, existence objects such as pedestrians and vehicles.

FIG. 3 illustrates vehicle surrounding environment information acquired by the vehicle surrounding information acquisition unit 121.
As shown in FIG. 3, the vehicle peripheral information acquisition unit 121 acquires a vehicle, a pedestrian, an obstacle, and other various information.
Specific information to be acquired includes, for example, the type of vehicle (passenger car, motorcycle, bicycle, etc.) existing in the detected vicinity, the inter-vehicle distance, the relative speed, and the attributes (oncoming vehicle, parallel vehicle, direct (left, right) ) Cars, etc. are acquired for each vehicle.
Similarly, information on each of pedestrians and obstacles is acquired.

The road information acquisition unit 122 includes a GPS device that detects the current position of the vehicle, and map information that acquires road information corresponding to the detected current position and peripheral information such as the presence or absence of a signal.
Further, the road information acquisition unit 122 includes an image recognition device that recognizes a sign and a road environment. This image recognition device is shared with the image recognition of the vehicle surrounding information acquisition unit 121.

FIG. 4 illustrates vehicle surrounding environment information acquired by the road information acquisition unit 122.
As shown in FIG. 4, in the road information acquisition unit 122, the road type, road shape, road width, own vehicle position, road surface condition, road brightness, presence / absence of signal, road attributes (traffic rules), other Various information is acquired.

The network unit 123 is connected to a traffic information network such as VICS and a weather information center, and acquires traffic information and weather information.
FIG. 5 illustrates vehicle surrounding environment information acquired by the network unit 123.
As shown in FIG. 5, the congestion information acquired by VICS includes the distance of congestion, the distance of congestion, the presence / absence of an accident, the presence / absence of traffic closure, the presence / absence of chain regulation, and the like.
In addition, the weather information acquired by the weather information center includes weather information such as sunny, cloudy, and rain, precipitation probability, temperature, and other information.

  The own vehicle surrounding environment information acquired by the own vehicle surrounding environment information acquisition unit 12 is a part of the own vehicle information acquired by the own vehicle information acquisition unit 11 (for example, straight ahead, right turn, left turn, etc. based on the amount of steering operation) The information is used for setting a situation according to a situation table 163 described later.

The biometric information acquisition unit 13 acquires biometric information for determining whether the driver during driving of the vehicle is in a normal state or an abnormal state, and as such a sensor, an electrocardiograph, a sphygmomanometer, a heart rate sensor, a perspiration sensor, or the like Various sensors are provided.
When the vehicle starts running, the biological information acquisition unit 13 detects the heart rate and the amount of sweat at predetermined time intervals and supplies the detected amount to the ECU 10.
For example, the heart rate sensor detects a heart rate by collecting a heart rate signal from the hand of a driver who is driving by an electrode disposed on a steering wheel. Note that the heart rate sensor may be provided with a dedicated sensor on the driver's body such as a wrist.
The sweat sensor is disposed on the steering and detects a sweat state from a change in a current value flowing depending on the sweat state.

FIG. 6 illustrates biometric information acquired by the biometric information acquisition unit 13.
In the biometric information acquisition unit 13, cardiac potential, RR interval, heart rate, respiratory rate, body temperature, blood pressure, skin potential, moisture loss (sweat volume), myoelectric potential, brain wave potential, and the like are acquisition targets.

The information providing unit 14 includes a driving operation assist unit, a voice output unit, and a screen output unit for performing a driving operation assist or warning according to the driving state of the driver.
FIG. 7 illustrates information provided by the information providing unit 14 and the contents of assist.
As shown in FIG. 7, the driving operation assist unit performs the steering operation assist, the accelerator operation assist, the brake operation assist, and the like as the assist for correcting the driving operation by the driver. Control. For example, when there is a wobble in the steering wheel operation by the driver, a torque operation is performed so that the steering becomes heavy, and when the braking force of the brake is weak, an assist is made so that the output with respect to the depression amount of the brake becomes large.
Further, the voice output unit outputs a warning voice and the screen output unit displays a warning screen according to the state of the driver.

The driver model processing unit 15 includes a driver model creation unit 151, a driver model storage unit 152, and a driver model output unit 153.
The driver model creation unit 151 functions as a driver model creation device, accumulates own vehicle information when the driver's state is normal among the own vehicle information acquired by the own vehicle information acquisition unit 11, and the normal A driver model is created from the vehicle information of the state.
The own vehicle information in the normal state is accumulated for each situation determined from the own vehicle surrounding environment information acquired by the own vehicle surrounding environment information acquisition unit 12 when the information is acquired, and a driver model is created for each situation. The

The driver model storage unit 152 stores the driver model created by the driver model creation unit 151 for each situation.
When a predetermined amount of the own vehicle information for each situation is accumulated, the driver model creation unit 151 creates a driver model for the situation and stores it in the driver model storage unit 152. Each time new own vehicle information is acquired, the driver model of the corresponding situation creates a new driver model together with the own vehicle information accumulated before then, and updates the driver model. The driver model may be updated or updated every time a predetermined amount of additional accumulation is made, not every time new vehicle information of the corresponding situation is acquired.

FIG. 8 conceptually shows the contents stored in the driver model storage unit 152.
As shown in FIG. 8, the driver model is classified for each situation. Each stored driver model a, b, c,... Is associated with corresponding situation data (situations a, b, c,...), And functions as a tag for the system to cite the driver model.
In this way, when searching for a driver model, it becomes possible to perform a cache operation such as acquiring driver models in a case of “the driver is fatigued at a certain level” in a batch.

Based on the driver model n corresponding to the specific situation n, the driver model output unit 153 estimates and outputs the driver's operation amount in the normal state, that is, the usual (normal) driving operation amount for the situation n. .
By comparing this estimated driving operation amount with the current host vehicle information, driving behavior deviation data, which is basic data for determining a driving behavior state (reaction delay, wobbling, etc.) described later, is a predetermined time interval. Obtained every time.

  Note that both functions of the driver model creation unit 151 and the driver model output unit 153 in the driver model processing unit 15 may be realized by the ECU 10, and the driver model storage unit 152 may be stored in the data storage unit 16.

The data storage unit 16 stores various data and tables necessary for the driver model creation process and the driving operation assist process according to the present embodiment.
The data storage unit 16 is optical information such as a magnetic recording medium such as a flexible disk, a hard disk, and a magnetic tape, a semiconductor recording medium such as a memory chip and an IC card, and a CD-ROM, MO, and PD (phase change rewritable optical disk). And other recording media on which data and computer programs are recorded by various methods.
Different recording media may be used depending on the recording contents.

The data storage unit 16 stores driving behavior deviation data 161 and host vehicle information 162, and also stores a situation table 163.
The driving behavior deviation data 161 is difference data between a normal driving operation amount estimated from the driver model n with respect to the current driving situation n and an operation amount based on actual own vehicle information. Is calculated and stored at predetermined time intervals for the situation n.
In the host vehicle information 162, host vehicle information when the vehicle travels in a normal state is accumulated for each situation. A driver model for the situation is created at the next point where the vehicle information is accumulated by a predetermined amount. Once the driver model is created, it is updated each time the own vehicle information of the corresponding situation is acquired.

The situation table 163 is a table for determining the corresponding situations a, b, c,... From the acquired own vehicle information and own vehicle surrounding environment information.
FIG. 9 conceptually shows the contents of situation data.
As shown in FIG. 9, a situation flag is set for each situation a, b, c,... Corresponding to the driver models a, b, c,.
As the situation flag, one data is selected for each small item in the host vehicle information and the host vehicle surrounding environment information.

Next, various processing operations performed by the driving support apparatus configured as described above will be described.
FIG. 10 is a flowchart showing the processing operation of the driver model creation process for creating a driver model of “normal driving behavior” (normal) of the driver.
In the present embodiment, the driver model is created while the vehicle is running. However, the driver's biometric information, own vehicle information, and surrounding environment information of the own vehicle are collected and accumulated during the running, and the setting of the situation flag and the driver are performed. The model may be created when the vehicle is not traveling.

The driver model creation unit 151 collects and accumulates biological information at each time point from the biological information acquisition unit 13 while the vehicle is traveling (step 10). Various information such as biological information is collected via the ECU 10 (the same applies hereinafter).
Next, the driver model creation unit 151 determines whether or not the current driver state is normal by monitoring the change state from the collected and accumulated biometric information (step 11).

11 to 13 conceptually show a method of determining whether or not the driver is in a normal state.
FIG. 11 shows a state in which a mental change due to fluctuation or impatience is monitored from fluctuations in the heart rate of the driver.
As shown in FIG. 11A, when the measured value of the heartbeat is between the predetermined upper and lower threshold values h1 and h2, it is determined that the state is normal (stable state).
On the other hand, as shown in FIG. 11B, when the measured value of the heartbeat is lower than the lower threshold value h1 or higher than the upper threshold value h2, an abnormal state (unstable state) due to shaking or impatience is detected. Judge.
In this embodiment, as shown in FIG. 11 (b), an abnormal state is determined when it deviates from both sides of the upper and lower thresholds h1 and h2 within a predetermined time, but either threshold is exceeded for a predetermined time. It may be determined that there is an abnormality.

FIG. 12 shows a state in which a mental change is monitored from Lorentz plot analysis of an electrocardiogram.
In Lorentz plot analysis, when the RR interval of the cardiac potential at an arbitrary time n is RRn and the RR interval of the cardiac potential at the next time n + 1 is RRn + 1, the horizontal axis is the value of RRn, and the vertical axis is A graph taking the value of RRn + 1 is created. Here, the RR interval is a time interval from the peak value of the electrocardiogram to the next peak value, and corresponds to the interval of heartbeats.

According to this Lorentz plot analysis, in the case of an extreme tension state, as shown in FIG. 12A, the heartbeat intervals are the same interval, and the set of plot points is concentrated at one place of the y = x line. .
Further, in a moderate tension state (a state with moderate attention), the heartbeat interval is observed with a moderate fluctuation, and as shown in FIG. 12B, the set of plot points is plotted elongated on the y = x line. The
When the attention is distracted, the heartbeat interval fluctuates, and as shown in FIG. 12C, the set of plot points also swells in the direction of the origin and the direction perpendicular to the y = x line. A large set is observed.
Further, in the state of drowsiness, as shown in FIG. 12 (d), the set of plot points has a heartbeat interval in which the plot area in the y = x-ray direction is widened, but the width on the origin side is narrow and the distance from the origin is increased. There is a tendency to become wide.

  By this Lorentz plot analysis, a normal state (moderate tension state) and an abnormal state (extreme tension state, state of distraction, and sleepiness) are determined.

FIG. 13 shows a case where it is determined from the acquired biological information whether the state is normal or not based on whether the sympathetic nervous system is dominant or the parasympathetic nervous system is dominant.
As shown in FIG. 13, for example, when the size of the pupil is measured from the captured image of the driver and the size is divergent, the exchange nervous system is the most dominant and attention is paid to the extreme tension state. It is judged that the power may be low. Conversely, if the pupil size is in a contracted state, the parasympathetic nervous system is in a relaxed state, and it is determined that the attention level may be low or the attention level may be very low depending on the degree of contraction. Is done.
On the other hand, in the case of a pupil size in which the exchange nervous system is moderately dominant, it is determined that the normal state is high in attention with moderate tension.

For each measurement item of action items (heart rate, cardiac contraction force, etc.) shown in FIG. 13 including the size of the pupil, extreme tension, moderate tension, relaxed state (low attention), relaxed state (caution) The value for classifying into four states (the force is extremely low) is determined in advance.
In creating the driver model, it is determined whether the driver model is in a normal state or not in a normal state. However, in the driver biometric information case determination (step 42) in the driver biometric information monitoring process (see FIG. 20) described later, an extreme tension state Five states are determined based on the method described with reference to FIGS. 11 to 13, a moderate tension state, a state where attention is diffused, a relaxation state, and a drowsiness state.

  As described above, it is determined whether or not the driver is in a normal state from the biometric information. If the driver is normal (step 11; Y), the driver model creating unit 151 uses the normal driver model information as normal driver model information. Vehicle information and host vehicle surrounding environment information are collected from the host vehicle information acquiring unit 11 and host vehicle surrounding environment information acquiring unit 12 (step 12).

FIG. 14 illustrates the host vehicle information and host vehicle surrounding environment information acquired by the host vehicle information acquiring unit 11 and host vehicle surrounding environment information acquiring unit 12 when making a right turn at an intersection.
When making a right turn at an intersection as shown in FIG. 14, the acquired information includes road type, road condition, own vehicle speed, own vehicle position, traveling direction, and signal state of the own vehicle (red, blue , Yellow, etc.), presence / absence of preceding vehicle, type of preceding vehicle, relative position of forward vehicle, relative speed of forward vehicle, presence / absence of oncoming vehicle, oncoming vehicle type, relative position of oncoming vehicle, relative speed of oncoming vehicle, presence / absence of pedestrians, walking The type of the person, the position of the pedestrian, the traveling direction of the pedestrian, the weather, etc. are acquired.
In this embodiment, these information is acquired and used to set the situation described later, but it is not always necessary to use all of them, and the situation may be set based on any part of the information. Conversely, situations may be set based on more detailed information.

  The driver model creation unit 151 sets a situation flag according to the situation table 163 (see FIG. 9) from the collected own vehicle information and the own vehicle surrounding environment information, and collects the own vehicle information at the normal time in the corresponding situation. The vehicle information 162 is accumulated (step 13).

  Next, the driver model creation unit 151 creates a normal driver model for normal operation corresponding to the situation set in step 13 in accordance with the collected and accumulated vehicle information 162 for normal operation (step 14), and returns to the main routine. To do.

  On the other hand, it is determined whether or not the driver is in a normal state from the biometric information. If the driver is not normal (step 11; N), the driver model creating unit 151 uses the abnormal driver model creation information as abnormal information as in the normal state. The own vehicle information and the own vehicle surrounding environment information are collected from the own vehicle information acquiring unit 11 and the own vehicle surrounding environment information acquiring unit 12 (step 15).

  The driver model creation unit 151 sets a situation flag according to the situation table 163 (see FIG. 9) from the collected own vehicle information and the surrounding environment information of the own vehicle, and collects the own vehicle information at the time of abnormality in the collected situation. The vehicle information 162 is accumulated (step 16).

  Next, the driver model creating unit 151 creates an abnormal driver model for an abnormal time corresponding to the situation set in Step 16 in accordance with the collected and accumulated own vehicle information 162 at the abnormal time (Step 17), and returns to the main routine. To do.

  In this embodiment, the case where the normal driver model and the abnormal driver model are created according to whether the biological information is normal or abnormal has been described. For example, as an abnormal state, the biological information is higher than a predetermined upper and lower threshold value. A driver model corresponding to the state of the biological information may be created so as to be low.

Here, creation of a driver model by the driver model creation unit 151 will be described.
In this embodiment, the driver model is created by GMM.
FIG. 15 shows the principle regarding the creation of a normal driver model and the estimation of the driving operation amount based on the created driver model by the driving support device (driver model creation device) in the present embodiment.
As for the creation of the driver model and the estimation of the driving behavior, the vehicle speed V, the inter-vehicle distance F with the preceding vehicle, and the first order dynamic feature amounts ΔV and ΔF (first-order differential values), second order Dynamic feature amounts ΔΔV and ΔΔF (second-order differential value), accelerator operation amount G and primary dynamic feature amount ΔG as a driver model for accelerator operation, and brake operation amount B and primary order as a driver model for brake operation. Although the case where the dynamic feature amount ΔB is used will be described, the feature amount may be created using a combination of other information among all pieces of information acquired by the own vehicle information acquisition unit 11. .

In the driving support device of this embodiment, the driving data 1 (own vehicle information) for each situation, which includes the accelerator operation amount, the vehicle speed, the inter-vehicle distance, etc., is used as learning data, and the driver model 2 by GMM corresponding to each situation is obtained. Prepared in advance by EM algorithm.
The driver model may be created for each driver.

When estimating the driver's driving behavior (for example, accelerator operation amount), the corresponding driver model 2 is used, and the maximum posterior probability for the measured values (V, F, ΔV,...) 3 of the travel data 1 at time t. By calculating 4, the accelerator operation amount 5 that the driver will operate is estimated.
Driving behavior deviation data is calculated by estimating the next operation amount using each operation amount estimated in this way as travel data and comparing it with an actual measurement value (vehicle information) at each time.

  In the driving support device of this example, based on the assumption that the driver determines the amount of operation of the accelerator pedal and the brake pedal based on the current vehicle speed, the inter-vehicle distance, and their primary and secondary dynamic feature amounts. ing.

Hereinafter, the principle of driver model creation and driving behavior estimation will be described in detail.
(A) Learning of Driver Model In the driver model 2 using the GMM, learning data is required, and traveling data 1 (own vehicle information) is used as a feature amount.
The travel data 1 uses time-series data for every predetermined measurement interval s (s is arbitrary, but s = 0.1 seconds in this embodiment).
The travel data 1 is data that is actually driven by the driver for which the driver model is created, and the travel data 1 measured and collected in real time when the driver is actually driving is used. Further, offline learning may be performed by using the travel data 1 measured and accumulated in advance.

In the driving support device according to the present embodiment, by creating a GMM for each driver, modeling that matches the characteristics of each driver is possible.
As described above, the driver model feature amount (running data 1) includes the vehicle speed, the inter-vehicle distance, and their primary and secondary dynamic feature amounts, the accelerator pedal operation amount, and the primary amount of the accelerator pedal operation amount. Of dynamic features are used.
In this way, by adding the dynamic feature amount to the feature amount and modeling, the time relationship before and after is taken into consideration, and a smooth and highly natural estimation result can be obtained.
In the description, the case where the primary and secondary dynamic feature quantities are used will be described. However, only the primary dynamic feature quantity may be used.

Similarly, a driver model for the operation amount of the brake pedal and other various operation amounts is also performed.
When creating a plurality of driver models for the accelerator pedal, the brake pedal, etc., the data other than the accelerator pedal operation amount, the brake pedal operation amount, etc. (V, F, ΔV, ΔF,...) Are the same data. May be used.

  In the present embodiment, the dynamic feature amount of the travel data 1 is calculated from the measured values of the accelerator operation amount, the vehicle speed, and the inter-vehicle distance, but may be actually measured.

In this embodiment, the driver model 2 is created by calculating a mixed Gaussian distribution (GMM) for the travel data 1.
That is, the joint probability density distribution for the running data 1 is calculated using the EM algorithm, and the parameters of the calculated joint probability density function = {λi, → μi, Σi | i = 1,2,3,. The driver model 2 is stored in a storage means such as a database.
Here, λi represents a weight, → μi represents an average vector group, Σi represents a variance-covariance matrix group, and M represents the number of mixtures. In addition, a symbol that is previously displayed as → μi means a vector.
As described above, the GMM according to the present embodiment uses the total covariance matrix in consideration of the correlation between the feature dimensions.

  As the EM algorithm, the mixed Gaussian distribution is estimated by the EM algorithm in accordance with, for example, Seiichi Nakagawa, “Voice Recognition by Probability Model” (The Institute of Electronics, Information and Communication Engineers 1988, P51 to P54).

(B) Estimation of driving behavior (various operation amounts such as accelerator pedal operation amount and brake pedal operation amount) The driver uses the accelerator pedal based on the current vehicle speed, the inter-vehicle distance, and their primary and secondary dynamic feature amounts. Based on the assumption that the operation amount of the brake pedal is determined, the movement behavior such as the operation amount of the pedal is estimated.
That is, from the simultaneous distribution of the feature amount, the motion behavior such as the accelerator pedal operation amount having the highest probability under the given condition is estimated.

This is a problem of maximizing conditional probability, and is based on the calculation of maximum posterior probability.
For example, the accelerator pedal operation amount ∧G (t) and the brake pedal operation amount ∧B (t) are estimates of the value x (t) that maximizes the conditional probability under the condition where y (t) is given. The maximum posterior probabilities are calculated by the following equations (1) and (2), respectively.

∧G (t) = arg max p (G | ΔG, V (t), F (t), ΔV (t), ΔF (t), ΔΔV (t), ΔΔF (t)) Equation (1)
∧B (t) = arg max p (B | ΔB, V (t), F (t), ΔV (t), ΔF (t), ΔΔV (t), ΔΔF (t)) Equation (2)

Here, like ∧G (t), the one displaying ∧ before means an estimated value.
Also,
p (G | ΔG, V, F, ΔV, ΔF, ΔΔV, ΔΔF)
= {P (G, V, F, ΔV, ΔF, ΔΔV, ΔΔF, ΔG)} / {∫∫ ... ∫p (G, V, F, ΔV, ΔF, ΔΔV, ΔΔF, ΔG) dΔG, dV, dF , dΔV, dΔF, dΔΔV, dΔΔF}
p (B | ΔB, V, F, ΔV, ΔF, ΔΔV, ΔΔF)
= {P (B, V, F, ΔV, ΔF, ΔΔV, ΔΔF, ΔB)} / {∫∫ ... ∫p (B, V, F, ΔV, ΔF, ΔΔV, ΔΔF, ΔB) dΔB, dV, dF , dΔV, dΔF, dΔΔV, dΔΔF}
It is.

  In equations (1) and (2), t represents time, and G, B, V, F, and Δ represent accelerator pedal operation amount, brake pedal operation amount, vehicle speed, inter-vehicle distance, and dynamic feature amount, respectively.

  However, the value of the accelerator pedal and the brake pedal that maximizes the conditional probability is obtained by performing numerical integration with a certain step size (for example, 100 steps from 0 to 10,000) in the interval from the minimum value to the maximum value. The value of the accelerator pedal and the brake pedal when the probability is calculated and the probability becomes the maximum may be used as the estimation result.

FIG. 16 shows an outline of driving behavior estimation based on the maximum posterior probability.
For simplicity, FIG. 16 shows a case where → x (t) is estimated when a feature value y (t) at a certain time t is given.

The driver driving behavior monitoring process for specifying the driving behavior state of the driver using the driver model created for each situation as described above will be described.
FIG. 17 is a flowchart showing the processing operation of the driver driving behavior monitoring process.
The ECU 10 collects host vehicle information and host vehicle surrounding environment information from the host vehicle information acquiring unit 11 and host vehicle surrounding environment information acquiring unit 12 (step 20).

Next, as shown in FIG. 18 (a), the ECU 10 sets a situation flag based on the acquired own vehicle information and own vehicle surrounding environment information (step 21).
Then, the ECU 10 performs matching processing with the situation table 163 based on the set situation flag, and searches for a situation that matches the current situation such as the acquired surrounding environment information of the host vehicle, thereby determining whether a corresponding driver model exists. Is determined (step 22).
If there is no corresponding driver model (step 22; N), the process returns to the main routine.

  On the other hand, as shown in FIG. 18B, when a situation that matches the current situation is searched and a corresponding driver model exists (step 22; Y), the ECU 10 associates the driver model linked with the matching situation. Is read from the driver model storage unit 152 and output to the driver model output unit 153 (step 23).

Next, the ECU 10 inputs the own vehicle information (actual value) acquired at the time t by the own vehicle information acquisition unit 11 to the driver model output unit 153 as an initial value (t) (step 24).
Then, the driver model output unit 153 inputs the own vehicle information (t) at time t to the driver model, and calculates the maximum posterior probability to obtain the estimated value “t + 1” of the driving action data (operation amount) at time t + 1. It outputs to ECU10 (step 25).

Next, the ECU 10 acquires the own vehicle information (t + 1) at the current time (time t + 1) (step 26), calculates driving behavior deviation data (“t + 1” − (t + 1)) at time t + 1, and outputs the driving behavior deviation data 161 as driving behavior deviation data 161. Store (step 27).
Then, the ECU 10 determines whether or not the stored driving behavior deviation data 161 has accumulated a predetermined number (step 28). If it is less than the predetermined number (step 28; N), the estimated value of the operation amount estimated in step 25 “T + 1” is input to the driver model as (t) (step 29), and the process proceeds to step 25 to further continue to accumulate driving behavior deviation data 161 at the next time (steps 25 to 27).

On the other hand, when the predetermined number of driving behavior deviation data 161 is accumulated (step 28; Y), the ECU 10 determines the driving behavior deviation tendency from the state of the driving behavior deviation data 161 and outputs it (step 30). Return to routine.
In the present embodiment, as the tendency to deviate from driving behavior, two items, the presence / absence of “reaction speed delay” and the presence / absence of “operation fluctuation”, are determined.

FIG. 19 conceptually compares the estimated value of the driving operation amount (normal driving) in the normal state output from the driver model output unit 153 with the operating amount of the current driving (own vehicle information). .
In FIG. 19, the driving operation amount (ordinary driving) output from the driver model has the highest probability as the driving behavior that the driver normally takes when the driver inputs the initial value of the current driving operation amount in the normal state. It represents the output value of the manipulated variable, and represents a virtual manipulated variable that should normally be operated (should be the manipulated variable) during normal operation.

The virtual operation amount is compared with the operation amount of the current driving to determine the tendency of reaction speed delay and operation fluctuation.
For example, as shown in FIG. 19A, when the operation amount estimated by the driver model increases with time, the operation amount based on the own vehicle information is acquired by the own vehicle information acquisition unit 11 after a predetermined time has elapsed. It is judged that there is a tendency for the reaction rate to be delayed.
Further, as shown in FIG. 19B, compared to the operation amount estimated by the driver model, the operation amount based on the acquired own vehicle information increases or decreases as time passes, or decreases. When the amount (absolute value of the driving behavior deviation data) is equal to or greater than a predetermined value, it is determined that the driving operation has fluctuation.

  On the other hand, when the operation amount estimated by the driver model and the operation amount based on the acquired own vehicle information substantially coincide with each other, that is, when the absolute value of the driving behavior deviation data continues to be a predetermined value or less, the reaction It is judged that there is no delay and no wobbling.

Next, the driver's biometric information monitoring process during traveling will be described with reference to the flowchart of FIG.
First, the ECU 10 collects and accumulates biological information at each time point from the biological information acquisition unit 13 while the vehicle is traveling (step 41).
Next, the ECU 10 monitors the change state from the collected and accumulated biometric information, and determines the current state (case) of the driver biometric information by the same method as the method described with reference to FIGS. 42) Return to the main routine.

The driver biometric information monitoring process described above includes the driver biometric information collection (step 10) in the driver model creation process during travel described in FIG. 10 and the determination of whether or not the change in the driver biometric information is normalized (step 11). ).
In this case, in the driver model creation process, it is monitored in step 11 that the driver is in the normal state, but as described with reference to FIG. It is also determined according to FIGS. 11 to 13 and output.

  In the case where the determination is also used, either the driver model creation unit 151 or the ECU 10 performs the determination.

Further, as shown in FIGS. 21A and 21B, the normal state, sleepiness and fatigue state may be determined from the eye state of the driver.
This determination from the eye state may be used in one or both of the driver model creation process (step 11) and the driver biometric information monitoring process (step 42).

Specifically, as shown in FIG. 21 (a), the number of blinks, the blink time, the eyelid opening, and the movement of the line of sight are detected by image processing as the driver state, and drowsiness is detected according to the value and state. Determine the state.
In addition, as shown in FIG. 21 (b), when the number of blinks is increased, when the movement of the eyelid is convulsed, when the eyes are rubbed, when the eyes are blurred, It is determined that

FIG. 22 is a flowchart showing the operation of the driving support process based on the driving state of the driver and the biological information case.
The ECU 10 obtains the driving behavior departure tendency determined and output in step 30 of the driver driving behavior monitoring process (FIG. 17) as the driver state (step 51) and the step of the driver biological information monitoring processing (FIG. 20). The driver biometric information case determined and output in 42 is acquired (step 52).
Then, the ECU 10 determines the content of driving support from the acquired driving behavior deviation tendency and the biological information case, approaches the driver (step 53), and returns to the main routine.

FIG. 23 is a table showing the driver state (a) estimated from the acquired driving behavior and the biometric information case, and the contents of driving assistance performed in accordance with the estimated driver state.
Note that the table of FIG. 23B is stored in the ROM of the ECU 10.
As shown in FIG. 23 (a), each driving action (reaction delay, wandering, reaction delay + wobbling, neither) and each biological information case (extreme tension state, moderate tension state, state of distraction) , Relaxed state, sleepy state), the driver's state is estimated such as being distracted, tired, casual driving, sleepiness, impatience, looking away, etc.
Then, corresponding to each state estimated from each of these combinations, as shown in FIG. 23 (b), the ECU 10 provides attention by voice or vibration, provides facility information, and recommends a break. The information providing unit 14 provides driving assistance such as providing information and distracting.

Of the driving assistance shown in FIG. 23, the ECU 10 controls + α1 because it concentrates on other than driving and is dangerous. Therefore, the ECU 10 urges attention and automatically increases the inter-vehicle distance from the preceding vehicle.
In addition, the ECU 10 alerts + α2 to concentrate on driving, and researches using the conversation function and sensors to determine what the driver feels so heavy that driving is negligible. To take a break to help them, get on a troubled consultation, and improve the problems the driver has.
In addition, the ECU 10 alerts + α3 to wake up and provides guidance such as taking a break immediately.
In addition, the ECU 10 may give a specific explanation for making it easy to obtain the driver's consent, such as a warning about what the driving operation or biometric information was, as a problem, as the content of the alert.

Although one embodiment of the driver model creation device and the driving support device of the present invention has been described above, the present invention is not limited to the described embodiment, and various modifications are made within the scope described in each claim. It is possible.
For example, in the described embodiment, the case where the content of the driving support is determined from the determined driving behavior and the biological information case has been described, but the driving support content may be determined from the determined driving behavior.

It is a block diagram of the driving assistance device to which the driver model creation device in one embodiment of the present invention is applied. It is explanatory drawing which illustrated the own vehicle information acquired by the own vehicle information acquisition part. It is explanatory drawing which illustrated vehicle surrounding environment information acquired in a vehicle surrounding information acquisition part. It is explanatory drawing which illustrated the vehicle surrounding environment information acquired in a road information acquisition part. It is explanatory drawing which illustrated the vehicle surrounding environment information acquired in a network part. It is explanatory drawing which illustrated the biometric information acquired in a biometric information acquisition part. It is explanatory drawing which illustrated the information provided in an information provision part, and the content of assistance. It is explanatory drawing which represented the memory content of the driver model memory | storage part notionally. It is explanatory drawing which represented the content of situation data notionally. It is a flowchart showing the processing operation of a driver model creation process for creating a driver model of “ordinary driving behavior” (normal) of a driver. It is explanatory drawing showing the state which monitors a mental change from the fluctuation | variation of the heart rate of a driver. It is explanatory drawing showing the state which monitors a mental change from the Lorentz plot analysis of an electrocardiogram. It is explanatory drawing showing the case where it is judged from the acquired biometric information whether it is a normal state. It is explanatory drawing which illustrated the own vehicle information acquired by the own vehicle information acquisition part and the own vehicle surrounding environment information acquisition part, and the own vehicle surrounding environment information when making a right turn at an intersection. It is explanatory drawing showing the principle regarding the production | generation of a driver model, and the estimation of the driving action based on the produced | generated driver model. It is explanatory drawing showing the outline regarding the estimation of the driving action by the maximum posterior probability. It is a flowchart showing the processing operation of the driver driving action monitoring process. It is explanatory drawing about the setting of a situation flag from the own vehicle information and own vehicle surrounding environment information, and the search of a suitable situation. It is explanatory drawing which compared notionally the estimated value of the driving operation amount (normal driving | operation) in the normal state output from the driver model output part, and the operating amount (own vehicle information) of the present driving | operation. It is a flowchart about the monitoring process of the biometric information of the driver during driving | running | working. It is explanatory drawing in the case of determining a normal state, drowsiness, and a fatigue state from the eye state of a driver. It is a flowchart showing operation | movement of the driving assistance process based on a driver driving | running state and a biometric information case. It is explanatory drawing which represented the content of the driving assistance performed corresponding to the state of the driver estimated from the acquired driving action and the biometric information case, and the estimated state of the driver.

Explanation of symbols

10 ECU
DESCRIPTION OF SYMBOLS 11 Own vehicle information acquisition part 12 Own vehicle surrounding environment information acquisition part 121 Vehicle periphery information acquisition part 122 Road information acquisition part 123 Network part 13 Biometric information acquisition part 14 Information provision part 15 Driver model process part 151 Driver model creation part 152 Driver model Storage unit 153 Driver model output unit 16 Data storage unit

Claims (7)

Driver model acquisition means for acquiring, as a driver model, a Gaussian mixture model obtained by calculating driving data including a plurality of types of feature amounts of driving operation in a normal state by an EM (Expectation Maximization) algorithm ;
Using the acquired driver model, driving operation estimation means for estimating a driving operation that is normally operated in a normal state;
Driving action determination means for determining the driving action of the driver from the estimated driving operation and the driving operation based on the current driving operation information;
Driving support means for providing driving support according to the determined driving behavior;
Comprising
The driving operation estimation means measures driving data excluding a specific feature quantity from among a plurality of types of feature quantities acquired by the driver model acquisition means, and calculates a maximum posterior probability in the driver model for the driving data. A driving support apparatus that estimates the specific feature amount .
The plurality of types of feature values of the driving operation acquired by the driver model acquisition means are travel data including a plurality of types of feature values of an accelerator operation amount, a vehicle speed, a brake operation amount, a steering operation amount, an inter-vehicle distance, and an acceleration. The driving support apparatus according to claim 1, wherein:
The driver model acquisition means acquires a driver model corresponding to the current driving environment from a driver model of a driving operation in a normal state created for each driving environment.
The driving support apparatus according to claim 1 , wherein the driving support apparatus is configured as described above.
Driver state determination means for determining the state of the driver from the driver's biometric information,
The driving support unit, the driving support apparatus according to claim 1, claim 2 or claim 3, characterized in that the driving support corresponding to the driver's state of the judgment and the judgment was driving behavior.
Driving support means, depending on the determination content, alert by voice or image, information provision, vibration,
Performing at least one or more driving support of the guide resting place, characterized in that claim 1,
The driving support device according to claim 2, claim 3, or claim 4 .
Driver model acquisition means for acquiring, as a driver model, a Gaussian mixture model obtained by calculating driving data including a plurality of types of feature amounts of driving operation in a normal state by an EM (Expectation Maximization) algorithm ;
Using the acquired driver model, driving operation estimation means for estimating a driving operation that is normally operated in a normal state;
Driving action determination means for determining the driving action of the driver from the estimated driving operation and the driving operation based on the current driving operation information;
Comprising
The driving operation estimation means measures driving data excluding a specific feature quantity from among a plurality of types of feature quantities acquired by the driver model acquisition means, and calculates a maximum posterior probability in the driver model for the driving data. A driving behavior determination device that estimates the specific feature amount .
The plurality of types of feature values of the driving operation acquired by the driver model acquisition means are travel data including a plurality of types of feature values of an accelerator operation amount, a vehicle speed, a brake operation amount, a steering operation amount, an inter-vehicle distance, and an acceleration. The driving behavior determination device according to claim 6.

Images