JP2011081570A - Driver condition estimating device and driver condition estimating method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a driver condition estimating device and a driver condition estimating method which can estimate a condition of a driver with high accuracy. <P>SOLUTION: An attention level estimating section 20 of an operation supporting device 10 acquires sample value groups of three parameters T<SB>N</SB>, H<SB>V</SB>and H<SB>R</SB>when an attention level of a driver is a low level or a high level, plots the groups on a vector space, makes a straight line N1 that connects points of average values in the sample value groups as a normal line, and sets a plane P1 that passes through a middle point M1 of the straight line N1. The attention level estimating section 20 estimates the attention level of the driver from a side of the plane P1 the measured T<SB>N</SB>, H<SB>V</SB>and H<SB>R</SB>are located at. As the sample value groups with different attention level of the driver are plotted on the three-dimensional vector space, it is easy to distinguish among the sample value groups with different attention level even if there is variation among the sample values of the sample value groups, so that is possible to estimate the condition of the driver with high accuracy. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a driver state estimation device and a driver state estimation method, and more particularly to a driver state estimation device and a driver state estimation method for estimating a driver state based on a sample value group of parameters influenced by the driver state. .

  In recent years, various information presentation methods using HMI (Human Machine Interface) have been studied as preventive safety technologies for the purpose of preventing traffic accidents in automobiles. In order to further improve the preventive safety effect by presenting information, in addition to the risk of physical accidents around the vehicle, consideration should be given to conditions such as the driver's caution level and fatigue level, which change every moment. Is considered important. For example, Patent Literature 1 discloses a fatigue detection device that estimates a driver's fatigue level by comparing a driver's operation amount such as a brake delay time with a reference operation amount. In this apparatus, when estimating the driver's fatigue level, estimation based on other vehicle information around the vehicle, road shape information, weather information, and operating device deterioration information is performed.

JP 2009-6839 A

  However, in the above-described technique, the driver's fatigue level is estimated by comparing the brake delay time, which is a single parameter, with the reference operation amount. The value of the driver's brake delay time and the like varies. For example, even if the driver is not tired at all, it is possible that an operation below a reference operation amount such as a preset brake delay time may occur due to various factors. In such a case, the device erroneously determines that the driver is tired. On the other hand, even if the driver is extremely tired, there is a possibility that the operation exceeds the reference operation amount such as a preset brake delay time. In such a case, the device erroneously determines that the driver is not tired. Even if other vehicle information, road shape information, etc. are simply taken into account in addition to the brake delay time, the above problem remains. Therefore, a method that can estimate the driver's state with higher accuracy is desired.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a driver state estimation device and a driver state estimation method capable of estimating a driver state with higher accuracy. .

  The present invention provides a sample of each of at least three parameters that are affected by the driver's state in each of the case where the driver's state is the first level and the case where the driver's state is a second level that is better than the first level. A sample value group acquiring means for acquiring a value group, a sample value group plotting means for plotting a sample value group of each parameter acquired by the sample value group acquiring means on a vector space having each parameter as a vector axis, and a sample value In the vector space where the sample value group of the parameter is plotted by the group plotting means, the first average value point that is the average value of each parameter in the first level sample value group, and the second level sample value group Mean value point plotting means for plotting a second average value point that is an average value of each parameter, and mean value point plotting means A straight line setting means for setting a straight line connecting the plotted first average value point and the second average value point, a midpoint setting means for setting a midpoint of the straight line set by the straight line setting means, and a straight line setting means A surface setting unit that sets a plane that passes through the midpoint of the straight line set by the midpoint setting unit, a measurement value acquisition unit that acquires a measured value of each parameter during the operation of the driver, and a measurement value acquisition The measured value plotting means for plotting the measured values of the respective parameters acquired by the means on the vector space, and the measured values plotted by the measured value plotting means are on the first level sample value side of the surface set by the surface setting means. When positioned, the driver estimates that the state is the first level, and the measured value plotted by the measured value plotting means is positioned on the second level sample value side of the surface set by the surface setting means. When is a driver state estimation device and a state estimation means for estimating a driver is in the state of the second level.

  According to this configuration, the sample value group acquisition unit affects the driver state in each of the case where the driver state is the first level and the case where the driver state is the second level better than the first level. The sample value group plotting means obtains the sample value group of each of the parameters obtained by the sample value group obtaining means and plots the sample value group of each parameter obtained by the sample value group obtaining means on the vector space having the parameters as vector axes. To do. In this way, by plotting sample value groups with different driver states on a three-dimensional or larger vector space, even if sample values in the sample value groups vary, sample value groups with different driver states It becomes easy to distinguish.

  Further, the mean value point plotting means is a first mean value point that is an average value of each parameter in the first level sample value group on the vector space where the sample value group of the parameter is plotted by the sample value group plotting means. And a second average value point that is an average value of each parameter in the second level sample value group, and the straight line setting means plots the first average value point and the second average value plotted by the average value point plotting means. A straight line connecting the value points is set, the midpoint setting means sets the midpoint of the straight line set by the straight line setting means, and the surface setting means sets the straight line set by the straight line setting means as the normal line and sets the midpoint By setting a plane that passes through the midpoint of the straight line set by the means, it is possible to set a plane that is a boundary between sample value groups with different driver states.

  Further, the measurement value acquisition means acquires the measurement value of each parameter during the driving of the driver, and the measurement value plotting means plots the measurement value of each parameter acquired by the measurement value acquisition means on the vector space to estimate the state. When the measurement value plotted by the measurement value plotting means is located on the first level sample value side of the surface set by the surface setting means, the means estimates that the driver is in the first level state, and the measurement When the measured value plotted by the value plotting unit is located on the second level sample value side of the surface set by the surface setting unit, the driver measures the parameter by estimating that it is in the second level state. Even when the values vary, it is possible to estimate the state of the driver with higher accuracy by using a surface that is a clear boundary.

In this case, if the parameter, which is the vehicle and the relative velocity ΔV of the preceding vehicle, the target following distance and the inter-vehicle distance error ΔR and the accelerator pedal displacement P _a is the difference between the present inter-vehicle distance, T N _(dP it is preferable to include _{a / dt) + P a =} H V ΔV + H R ΔR pedal delay time constant _{T N} in the accelerator pedal operation model represented by the relative speed feedback gain _{H V} and the inter-vehicle distance error feedback gain _{H R.}

According to this arrangement, the pedal delay time constant T _N in the accelerator pedal operation model when the vehicle is performing a follow-up running against a preceding _vehicle, a relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R parameter Therefore, the driver's state can be estimated with higher accuracy during the follow-up driving in which the state such as the driver's attention level is important.

  Further, the sample value group acquisition means is provided when the driver state is the first level and when the driver state is from the first level to the nth level (3 <n) in which the driver state is improved stepwise from the first level. The sample value group for each parameter is acquired, and the mean value point plotting means plots the first mean value point to the nth mean value that is the mean value of each parameter in the sample value group for each of the first level to the nth level. The straight line setting means connects the first straight line to the (n-1) th average value point and the nth average value point connecting the first average value point and the second average value point plotted by the average value point plotting means. The (n-1) th straight lines are respectively set, and the midpoint setting means sets the first midpoint to the (n-1) th midpoint of each of the first to (n-1) th straight lines set by the straight line setting means. The surface setting means modulo the first to (n-1) th straight lines set by the straight line setting means. The first to n-1th planes passing through the midpoints of the 1st to n-1th straight lines set by the midpoint setting means are respectively set, and the state estimating means Is the measurement value plotted by the measurement value plotting means with respect to the kth level sample value of the kth plane in the positional relationship with the kth plane (1 ≦ k <n−1) closest to the measurement value. When the position is on the side, the driver estimates that the state is the k-th level, and the measurement value plotted by the measurement value plotting means is the side of the k + 1-th level sample value of the k-th plane set by the plane setting means. It is preferable to estimate that the driver is in the state of the (k + 1) th level.

  According to this configuration, the sample value group acquisition means includes the first level to the nth level (3 <n) when the driver state is the first level and when the driver state is gradually improved from the first level. In order to obtain the sample value group of each parameter in the case of, the sample value group of the driver state classified into three or more stages is plotted on the vector space.

  The mean value point plotting means plots the first mean value point to the nth mean value, which is the mean value of each parameter in the sample value groups of the first level to the nth level, and the straight line setting means The first straight line connecting the first average value point and the second average value point plotted by the value point plotting means to the (n-1) th straight line connecting the n-1 average value point and the nth average value point, respectively. The midpoint setting means sets the first midpoint to the (n-1) th midpoint of each of the first to (n-1) th straight lines set by the straight line setting means, and the plane setting means sets the straight line. The first to (n-1) th straight lines set by the means are normals, and the first through the middle points of the first to (n-1) th straight lines set by the midpoint setting means are the first ones. Sample values with different driver states classified into three or more levels on the vector space in order to set the plane to the (n-1) th plane. It is possible to demarcate the surface from the other.

  Further, the state estimating means is the kth surface of the kth surface in the positional relationship between the measurement value plotted by the measurement value plotting device and the kth surface (1 ≦ k <n−1) closest to the measurement value. When located on the sample value side of the level, the driver estimates that the state is the k-th level, and the measurement value plotted by the measurement value plotting means is the k + 1th level of the k-th plane set by the plane setting means. When it is located on the sample value side, it is estimated that the driver is in the state of the (k + 1) th level, so that the driver's state can be estimated in more than three stages.

  Further, it is preferable that the plane setting means sets a plane passing through the midpoint of the straight line set by the midpoint setting means with the straight line set by the straight line setting means as a normal line.

  According to this configuration, the surface setting means sets the plane passing through the midpoint of the straight line set by the midpoint setting means with the straight line set by the straight line setting means as the normal line, and therefore the sample value groups with different driver states It is possible to easily set a surface that becomes a boundary between each other.

  On the other hand, the surface setting means sets a curved surface that passes between the sample value groups with different driver states, using the straight line set by the straight line setting means as a normal line, passing through the midpoint of the straight line set by the midpoint setting means. It is preferable to do.

  According to this configuration, the surface setting means uses the straight line set by the straight line setting means as a normal, passes through the midpoint of the straight line set by the midpoint setting means, and passes between sample value groups having different driver states. In order to set a curved surface that passes through, for example, even when there is a large variation in sample values in the sample value group, it is possible to reliably set a surface that is a boundary between sample value groups with different driver states.

  On the other hand, according to the present invention, each of at least three parameters that are affected by the driver state in each of the case where the driver state is the first level and the case where the driver state is the second level better than the first level. A sample value group acquisition step of acquiring a sample value group of the sample, a sample value group plotting step of plotting the sample value group of each parameter acquired in the sample value group acquisition step on a vector space having the parameters as vector axes, On the vector space where the sample value groups of the parameters are plotted by the sample value group plotting step, the first average value points that are the average values of the parameters in the first level sample value groups and the second level sample value groups An average value point plotting process for plotting the second average value point, which is the average value of each parameter, and an average value point plot A straight line setting step for setting a straight line connecting the first average value point and the second average value point plotted in the step, a midpoint setting step for setting a midpoint of the straight line set in the straight line setting step, and a straight line setting step. The surface setting process that sets the plane that passes through the midpoint of the straight line set by the midpoint setting process, the measurement value acquisition process that acquires the measured values of each parameter during the driver's operation, and the measurement The measured value plotting step for plotting the measured values of each parameter acquired in the value acquiring step on the vector space, and the measured values plotted in the measured value plotting step are the first level sample values of the surface set in the surface setting step. When it is located on the side, the driver estimates that the state is the first level, and the measured value plotted in the measured value plotting process is on the second level sample value side of the surface set by the surface setting process. When location, the driver is in the driver state estimation method comprising the state estimation step of estimating that the state of the second level.

In this case, if the parameter, which is the vehicle and the relative velocity ΔV of the preceding vehicle, the target following distance and the inter-vehicle distance error ΔR and the accelerator pedal displacement P _a is the difference between the present inter-vehicle distance, T N _(dP it is preferable to include _{a / dt) + P a =} H V ΔV + H R ΔR pedal delay time constant _{T N} in the accelerator pedal operation model represented by the relative speed feedback gain _{H V} and the inter-vehicle distance error feedback gain _{H R.}

  In the sample value group acquisition step, the driver state is the first level and the driver state is the first level to the nth level (3 <n) in which the driver state is improved stepwise from the first level. The sample value group for each parameter is acquired, and in the mean value point plotting step, the first mean value point to the nth mean value that are the mean values of the parameters in the sample value group for each of the first level to the nth level are plotted. In the straight line setting step, the first straight line connecting the first average value point and the second average value point plotted in the average value point plotting step to the n-1th average value point and the nth average value point are obtained. The n-1th straight lines to be connected are set, and in the midpoint setting step, the first midpoint to the n-1th midpoint of each of the first to n-1 straight lines set in the straight line setting step are set. In the surface setting process, the first to nth set in the straight line setting process The first surface to the (n-1) th plane passing through the midpoints of the first to the (n-1) th straight lines set in the midpoint setting step, with the first straight line as a normal line, respectively. In the state estimation step, the measurement value plotted in the measurement value plotting step is the position of the kth surface in the positional relationship with the kth surface (1 ≦ k <n−1) closest to the measurement value. When located on the k-level sample value side, the driver estimates that the driver is in the k-th level, and the measurement value plotted in the measurement value plotting process is the (k + 1) th of the k-th plane set in the plane setting process. When the level is on the sample value side, it is preferable to estimate that the driver is in the (k + 1) th level.

  In the surface setting step, it is preferable to set a plane passing through the midpoint of the straight line set in the midpoint setting step with the straight line set in the straight line setting step as a normal.

  On the other hand, in the surface setting process, the straight line set in the straight line setting process is used as a normal, passing through the midpoint of the straight line set in the midpoint setting process, and a curved surface passing between sample value groups with different driver states is set. It is preferable to do.

  According to the driver state estimation device and the driver state estimation method of the present invention, it is possible to estimate the driver state with higher accuracy.

It is a figure which shows the structure of the driving assistance device which concerns on embodiment. It is a flowchart which shows operation | movement of the driving assistance device of embodiment. It is a side view showing the situation where the driving support device of an embodiment is applied. It is a perspective view which shows the three-dimensional space which set the plane for estimating the driver's attention level. It is a perspective view which shows the three-dimensional space which set the curved surface for determining the driver's attention level. It is a perspective view which shows the three-dimensional space which set the some plane for estimating the driver's attention level. It is a table | surface which shows the test subject in an experiment example. It is a figure which shows the subtask in an experiment example. It is a graph which shows the questionnaire result of the subjective attention level with respect to a test subject's driving | operation at the time of confirming the isolation | separation degree of the sample value group by the height of a test subject's attention level first in an experiment example. It is a table | surface which shows the degree which the level of attention level was able to isolate | separate when confirming the separation degree of the sample value group by the level of a subject's attention level first in an experiment example. It is a perspective view which shows the three-dimensional space which set the plane for estimating the attention level at the time of confirming the isolation | separation degree of the sample value group by the level of a subject's attention level first in an experiment example. It is a graph which shows the fluctuation | variation of the inter-vehicle distance, speed, and pedal operation amount when there is no subtask when the subject's attention level is estimated for the second time in the experimental example. It is a graph which shows the change of the inter-vehicle distance, speed, and pedal operation amount when there is a subtask when the subject's attention level is estimated for the second time in the experimental example. It is a perspective view which shows the three-dimensional space which set the plane for estimating the attention level at the time of estimating a test subject's attention level the 2nd time in an experiment example. It is a table | surface which shows the degree which was able to divide the presence or absence of a subtask, when a test subject's attention level was estimated the 2nd time in the experiment example. It is a graph which shows the questionnaire result of the subjective attention level with respect to a test subject's driving | operation at the time of estimating a test subject's attention level the 2nd time in an experiment example.

  Hereinafter, a driver state estimation device and a driver state estimation method according to an embodiment of the present invention will be described with reference to the drawings. A driving support device 10 shown in FIG. 1 is a device that is mounted on a vehicle and performs driving support such as presentation of information according to a driver's attention level.

  As shown in FIG. 1, the driving support device 10 includes a radar sensor 11, a camera sensor 12, a pedal sensor 13, and a vehicle speed sensor 14. The radar sensor 11 is for measuring the distance between the host vehicle and the preceding vehicle and the speed of the preceding vehicle. As the radar sensor 11, a millimeter wave radar or a laser radar can be applied. The camera sensor 12 is also for measuring the inter-vehicle distance between the own vehicle and the preceding vehicle. A stereo camera or the like can be applied to the camera sensor 12. The pedal sensor 13 is for detecting the displacement of the accelerator pedal of the host vehicle. The vehicle speed sensor 14 is for detecting the vehicle speed of the own vehicle. The vehicle speed detected by the vehicle speed sensor 14 is used for correcting the detection values detected by the radar sensor 11, the camera sensor 12, and the pedal sensor 13 in addition to estimating the driver's attention level.

  The driving support device 10 includes a caution level estimation unit 20. The attention level estimation unit 20 is for estimating the driver's attention level based on detection values detected by the radar sensor 11, the camera sensor 12, the pedal sensor 13, and the vehicle speed sensor 14. The attention level estimation unit 20 is physically composed of an ECU (Electro Control Unit) and controls the entire apparatus.

  A driver database 30 is connected to the attention level estimation unit 20. In the driver database 30, sample values of parameters for the attention level estimation unit 20 to estimate the driver's attention level are recorded. The sample values recorded in the database 30 can be sample values specific to the driver by using data for each driver of the vehicle. Alternatively, when there is a general tendency that does not depend on individual drivers in the parameter values for the driver's attention level, the sample values recorded in the database 30 should be statistical sample values for general drivers. You can also. In this case, sample values that match the age, sex, driving experience, etc. of the driver of the vehicle can be applied.

  The assist method determination unit 40 is configured to monitor the environmental risk that the vehicle estimated based on the detection values detected by the radar sensor 11, the camera sensor 12, the pedal sensor 13, and the vehicle speed sensor 14 is placed in the current environment, The display 51, the speaker 52, the buzzer 53, and the reaction force actuator 54 are operated according to the driver's attention level estimated by the level estimation unit 20 to correct the driver's attention level and assist the driver's driving. Is.

Hereinafter, operation | movement of the driving assistance device 10 of this embodiment is demonstrated. As a whole, the driving support device 10 of the present embodiment repeatedly performs the operation shown in FIG. 2 while the vehicle is traveling. In this embodiment, as shown in FIG. 3, it is assumed that the host vehicle 100 is traveling following the preceding vehicle 200. As shown in FIG. 2, the radar sensor 11, the camera sensor 12, the pedal sensor 13, and the vehicle speed sensor 14 acquire each data (S11). In this case, as shown in FIG. 3, the vehicle speed V of the vehicle 100, the preceding vehicle velocity V _P and the inter-vehicle distance R of the preceding vehicle 200 is detected.

The attention level estimation unit 20 is a model for estimating a driver's attention level from detection values detected by the radar sensor 11, the camera sensor 12, the pedal sensor 13, and the vehicle speed sensor 14 or data recorded in the driver database 30. Are identified (S12). In the present embodiment, the accelerator pedal operation model for estimating the attention level is expressed by the following equation (1).

P _a is the accelerator pedal displacement of formula (1), [Delta] V is the relative velocity, [Delta] R is the inter-vehicle distance error, T _N is the pedal delay time constant, H _V is the relative velocity feedback gain, H _R is the inter-vehicle distance error feedback gain. The inter-vehicle distance error ΔR is a difference between the target inter-vehicle distance and the current inter-vehicle distance R. The pedal delay time constant _TN is a constant representing the degree to which the accelerator pedal is depressed gently. The relative velocity feedback gain H _V, is a constant indicating the degree of speed of response relative speed is modified. The inter-vehicle distance error feedback gain H _R, is a constant indicating the degree of speed of response following distance error ΔR is corrected.

  The attention level estimation unit 20 estimates the driver's attention level for the above accelerator pedal operation model by a method described later (S13). The assist method determination unit 40 estimates an environmental risk in which the vehicle is estimated in the current environment, which is estimated based on detection values detected by the radar sensor 11, the camera sensor 12, the pedal sensor 13, and the vehicle speed sensor 14. (S14).

  The assist method determination unit 40 determines whether to perform assist according to the environmental risk and the driver's attention level estimated by the attention level estimation unit 20 (S15). When assist is to be executed, the assist method determination unit 40 performs assist such as warning by operating the display 51, the speaker 52, the buzzer 53, and the reaction force actuator 54 (S16). In this case, the driver is driven by the reaction force applied to the accelerator pedal of the reaction force actuator 54 by the display 51, the voice guidance by the speaker 52, the alarm notification by the buzzer 53, and the descending order of the environmental risk and the driver's attention level. It is done with increased driving intervention and intensity.

Hereinafter, a method for estimating the driver's attention level in the present embodiment will be described. The attention level estimation unit 20 determines whether the driver's attention level is a low attention level or a high attention level based on the detection value detected by the radar sensor 11 or the like or the data recorded in the driver database 30. in acquires the pedal delay time constant T _N, the respective sample values three parameters relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R.

As shown in FIG. 4, note the level estimation unit 20, the pedal delay time constant T _N, the three parameters each sample value group of the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R, these parameters are Plot on the vector space with x as the vector axis. As shown in FIG. 4, the pedal delay time constant T _N, the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H 3-dimensional vector space to _R 3 axes, respectively shown by the white triangles plot and white circles plot The low attention level sample value group and the high attention level sample value group are plotted.

Note level estimating unit 20 is a low attention level and high attention level sample value group of each pedal delay time constant T _N, the point at which the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R respectively mean Plot. As shown in FIG. 4, on the three-dimensional vector space, an average point with a low attention level and an average point with a high attention level, which are respectively indicated by a black triangle plot and a black circle plot, are plotted.

  As shown in FIG. 4, the attention level estimation unit 20 sets a straight line N1 that connects the average point of the low attention level and the average point of the high attention level. As shown in FIG. 4, the attention level estimation unit 20 sets a midpoint M1 of the straight line N1. As shown in FIG. 4, the attention level estimation unit 20 sets a plane P1 passing through the midpoint M1 of the straight line N1 with the straight line N1 as a normal line. Thereby, the plane P1 that divides the sample value group of the low attention level and the high attention level is set.

Then, the radar sensor 11, the camera sensor 12, the detected value of the pedal sensor 13 and the vehicle speed sensor 14, the pedal delay time constant T _N, the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R of each value obtained In such a case, the attention level estimation unit 20 plots them on a three-dimensional space, and when the plotted point is located on the low attention level side of the plane P1, it is estimated that the driver has a low attention level. When the plotted point is located on the high attention level side of the plane P1, the driver estimates that the attention level is high.

  In this case, the surface that divides the sample value group of the low attention level and the high attention level is not limited to a plane, and may be a curved surface P1 'as shown in FIG. In this case, the curved surface P1 'passes through the midpoint M1 of the straight line N1 with the straight line N1 as a normal line, as in the plane P1 of FIG. The curved surface P1 'is a curved surface that passes between a sample value group having a low attention level and a high attention level. In this case, as shown in FIG. 5, the attention level estimation unit 20 uses the sample values in the sample values group with the low attention level and the sample values with the closest distance between the sample values in the sample value group with the high attention level. A straight line N1 ′ connecting is established. Further, the midpoint M1 'of the straight line N1' is set similarly. The attention level estimation unit 20 sets the curved surface P1 'as a curved surface passing through the midpoint M1' of the straight line N1 'with the straight line N1' as a normal. As a result, a curved surface P <b> 1 ′ that passes between the sample values of the low attention level and the high attention level is set.

Alternatively, in this case, as shown in FIG. 6, three or more levels of attention may be set, and for each of the sample value groups, a surface serving as a boundary between the sample value groups may be set similarly. In the example of FIG. 6, average values are similarly obtained for the sample value groups of the first to fourth attention levels, straight lines N1 to N3 and their midpoints M1 to M3 are set, and the straight lines N1 to N3 are normal lines. And planes P1 to P3 passing through the midpoints M1 to M3 of the straight lines N1 to N3 are set. In this case, caution level estimation unit 20, the radar sensor 11, the camera sensor 12, the detected value of the pedal sensor 13 and the vehicle speed sensor 14, the pedal delay time constant T _N, the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _When each value of _R is obtained, the driver's attention is determined by which attention level the sample value group is located on the side of the sample value group on the planes P1 to P3 closest to the point on which the value is plotted. Estimate the level.

  According to the present embodiment, the attention level estimation unit 20 samples each of at least three parameters affected by the driver's attention level when the driver's attention level is low and when the driver's attention level is high. A value group is acquired, and a sample value group of each parameter is plotted on a vector space having each parameter as a vector axis. Thus, by plotting sample value groups with different driver attention levels on a three-dimensional or larger vector space, even if sample values in the sample value groups vary, the sample value groups with different driver attention levels It becomes easy to distinguish each other.

  In addition, the attention level estimation unit 20 plots the average value points that are the average values of the parameters in the sample value groups of the low attention level and the high attention level on the vector space where the sample value groups of the parameters are plotted. By setting the straight line N1 connecting the average value points, setting the midpoint M1 of the straight line, setting the straight line N1 as the normal line, and setting the plane P1 passing through the midpoint M1 of the straight line, the driver's attention A plane serving as a boundary between sample value groups having different levels can be set.

  Further, the attention level estimation unit 20 acquires the measured values of each parameter during the driving of the driver, plots the measured values of each parameter on the vector space, and the plotted measured value is a low attention level of the set plane P1. When the measured value is located on the side of the high attention level sample value on the plane P1, the driver is assumed to have a high attention level. By estimating the state, it is possible to more accurately estimate the driver's attention level using the plane P1 that is a clear boundary even when the measured values of the parameters vary.

Further, in the present embodiment, the pedal delay time constant T _N in the accelerator pedal operation model when the vehicle 100 is performing follow-up running against the preceding vehicle _200, the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _{Since R} is included in the parameter, the driver's attention level can be estimated with higher accuracy during the follow-up driving in which the driver's attention level is important.

  In the present embodiment, the attention level estimation unit 20 acquires the sample value group of each parameter when the driver's attention level is the first level to the fourth level. Attention level sample values will be plotted on the vector space.

  In addition, the attention level estimation unit 20 plots average value points that are average values of the parameters in the sample value groups of the first level to the fourth level, and straight lines N1 to N3 that connect these average value points to each other. The midpoints M1 to M3 are set, the straight lines N1 to N3 are normal, and the planes P1 to P3 passing through the midpoints M1 to M3 of the straight lines N1 to N3 are set. For this reason, it is possible to set a plane serving as a boundary between sample value groups having different driver attention levels classified into three or more stages on the vector space.

  Furthermore, the attention level estimation unit 20 determines which attention level the sample value group is on the side of the sample value group in which the plotted measurement value is in the positional relationship with the planes P1 to P3 closest to the measurement value. Estimate the driver's attention level. Therefore, it is possible to estimate the driver state in more than three stages.

  In the present embodiment, the attention level estimation unit 20 sets the plane P1 that passes through the midpoint M1 of the straight line N1 with the set straight line N1 as a normal line, and thus a surface that is a boundary between sample value groups with different driver attention levels. Can be set easily. Alternatively, according to the present embodiment, the attention level estimation unit 20 uses the set straight line N1 as a normal, passes through the midpoint M1 of the straight line N1, and the curved surface P1 passing between sample value groups having different driver attention levels. Since 'is set, for example, even when there is a large variation in the sample values in the sample value group, it is possible to reliably set the surface serving as the boundary between the sample value groups having different driver attention levels.

(Experimental example)
Hereinafter, experimental examples of the present invention will be described. Driver model pedal delay time constant T _N is a parameter of the equation _(1), the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R, it was identified by the least squares method using the traveling data for the last 2 minutes . Calculation was performed online using a driving simulator (hereinafter referred to as DS), and the calculation was performed every 10 seconds.

In order to estimate the attention level based on the parameters of the driver model, the driving data with high and low attention levels were acquired and the parameters were identified. Let these identified parameters be normative parameters at each level of attention. The standard parameter T _N of driver model to estimate the caution _level, obtains the H _V and H _R, to determine the plane P1 as the threshold of the attention level estimation shown in FIG.

  In this experimental example, three males in their 20s who had a driving license were used as subjects. FIG. 7 shows the age, driving history, and annual traveling distance of each subject. In the experiment, DS with a shaker (manufactured by Mitsubishi Precision Corporation) was used from the viewpoint of ensuring the safety of the subject and reproducibility of the experimental conditions.

  In order to simulate a state of low attention level, subjects were asked to perform subtasks. The subtask is to present one-digit numbers from the speaker continuously at intervals of 3 seconds, and cause the subject to add one-digit numbers and make verbal responses. For example, as shown in FIG. 8, when the numbers presented by voice are 5, 7, 1, 3, the test subjects are (5 + 7 =) 12, (7 + 1 =) 8, (1 + 3 =) 4 every 3 seconds. I will answer.

The preceding vehicle following scene was set on DS as an experimental scenario. The driving course is the left lane of the two-lane highway. The preceding vehicle accelerates / decelerates at random timing, and travels at a traveling speed of 70, 80, or 90 km / h. The acceleration / deceleration at this time is 1 m / s ² . At a random timing between 5 and 8 minutes from the start of travel, the preceding vehicle suddenly decelerates at 7 m / s ² , stops as it is, and the experiment ends. The subject explained the driving scenario in advance, and the experiment was conducted after enough practice driving to get used to the driving operation in DS. During driving, the subjects were instructed to keep the relative speed small and to maintain an appropriate inter-vehicle distance according to the speed. Travel data for a total of 30 trips was collected for each test subject, with 5 trips each when the subtask was performed and when the subtask was not performed. Each time the subject traveled one trip, the level of attention to driving was evaluated in 11 steps from 0 to 10 as a questionnaire.

  FIG. 9 shows the results of a questionnaire conducted by the subject for each trip. In FIG. 9, E1 to E3 represent three subjects. As a result of comparing the level of subjective attention level with and without subtasks, differences in subjective attention levels were 5%, 1%, and 1% in subjects E1 to E3, respectively.

  FIG. 10 shows a table in which parameters are identified from the running data of each subject obtained through experiments, and how much the level of attention level can be separated by the plane for estimating the attention level determined thereby. The degree of separation is determined from the positional relationship between the parameter group and the plane. For example, when there are 10 parameter points when there is no subtask and 8 of them are on the side representing a higher attention level than the plane, the separation accuracy is 80%. Referring to FIG. 10, it can be confirmed that the subjects E1 to E3 can separate the attention level from 70 to 80%.

  FIG. 11 shows a parameter identification result of the subject E1 and a plane P1 for estimating the attention level. Further, the inter-vehicle distance, travel speed, and accelerator pedal operation amount with and without subtasks are shown in FIG. 12 for E1 without subtasks, and in FIG. 13 with subtasks, respectively. Comparing FIG. 12 and FIG. 13, it appears that the followability with respect to the speed of the preceding vehicle is reduced as a result of the subtask, and the operation frequency of the accelerator pedal is reduced. Similarly, E2 and E3 have an influence on driving operation behavior.

  Next, a verification experiment of estimation accuracy was performed using the plane P1 determined in the above-described experiment. Under the same conditions as the above-described experiment, a total of 30 trip experiments were performed with and without subtasks. Parameter identification was performed online at DS.

  FIG. 14 shows an identification result of the subject E1 and a plane P1 for attention level estimation. FIG. 15 shows the estimation results by the plane of each subject, divided by the presence or absence of subtasks. From FIG. 15, it can be confirmed that the estimation accuracy is 60 to 90%.

  FIG. 16 shows the results of a questionnaire conducted in Experiment 2. The subjects E1 and E3 had a significant difference as in FIG. In the present invention, an attention level estimation method using a driver model was examined in order to improve the effect of preventive safety by HMI for information presentation. In order to estimate the attention level, a plane for identification was set on a space having parameters on three axes, and the estimation accuracy of the attention level was investigated. As a result, it was shown that the level of driver's attention level can be identified with 60 to 90% accuracy by the proposed method.

It should be noted that the present invention is not limited to the above-described embodiment, and it is needless to say that various modifications can be made without departing from the gist of the present invention. For example, in the above embodiment, the method for estimating the driver's attention level as the driver state has been described. However, the present invention is not limited to this, and various states such as the driver's fatigue level and arousal level can be set. It is applicable when estimating. Further, the above-described embodiment, as parameters used for the estimation, the pedal delay time constant T _N, but using the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R, can also be used other parameters, the parameters The number is not limited to three, and four or more parameters can be used.

DESCRIPTION OF SYMBOLS 10 ... Driving assistance device, 11 ... Radar sensor, 12 ... Camera sensor, 13 ... Pedal sensor, 14 ... Vehicle speed sensor, 20 ... Attention level estimation part, 30 ... Driver database, 40 ... Assist method determination part, 51 ... Display, 52 ... Speaker, 53 ... Buzzer, 54 ... Reaction force actuator, 100 ... Own car, 200 ... Preceding car.

Claims (10)

Sample values of each of at least three parameters affected by the driver state in each of the case where the driver state is the first level and the case where the driver state is a second level better than the first level. Sample value group acquisition means for acquiring a group;
Sample value group plotting means for plotting the sample value group of each of the parameters acquired by the sample value group acquiring means on a vector space with each of the parameters as a vector axis;
On the vector space where the sample value group of the parameter is plotted by the sample value group plotting means, a first average value point that is an average value of each of the parameters in the sample value group of the first level; Average value point plotting means for plotting a second average value point that is an average value of each of the parameters in the sample value group of the second level;
Straight line setting means for setting a straight line connecting the first average value point and the second average value point plotted by the average value point plotting means;
A midpoint setting means for setting a midpoint of the straight line set by the straight line setting means;
Surface setting means for setting the surface passing through the midpoint of the straight line set by the midpoint setting means, with the straight line set by the straight line setting means as a normal line,
Measurement value acquisition means for acquiring measurement values of each of the parameters during operation of the driver;
Measurement value plotting means for plotting the measurement values of the parameters acquired by the measurement value acquisition means on the vector space;
When the measured value plotted by the measured value plotting unit is located on the sample value side of the first level of the surface set by the surface setting unit, the driver is in the first level state. And when the measured value plotted by the measured value plotting means is positioned on the sample value side of the second level of the surface set by the surface setting means, the driver State estimation means for estimating that the state of
A driver state estimating device. If the parameter was the vehicle and the relative velocity ΔV of the preceding vehicle, the target following distance and the inter-vehicle distance error ΔR and the accelerator pedal displacement P _a is the difference between the present inter-vehicle distance,
T _N (dP _a / dt) + P _a = H _V ΔV + H _R ΔR
In pedal delay in the accelerator pedal operation model shown time constant T _N, including the relative velocity feedback gain H _V and the inter-vehicle distance error feedback gain H _R, the driver condition estimating apparatus according to claim 1. The sample value group acquisition means has the first level to the nth level (3 <n) when the driver state is the first level and when the driver state is gradually improved from the first level. A sample value group for each of the parameters in the case,
The average value point plotting means plots the first average value point to the n-th average value that is the average value of each of the parameters in the sample value group of the first level to the n-th level,
The straight line setting means includes a first straight line connecting the first average value point plotted by the average value point plotting means and the second average value point to the (n-1) th average value point and the nth average value. Set each of the (n-1) th straight lines connecting the points,
The midpoint setting means sets the first midpoint to the n-1th midpoint of each of the first to n-1 straight lines set by the straight line setting means,
The surface setting means uses the first to n-1 straight lines set by the straight line setting means as normals, and the first to n-1 straight lines set by the midpoint setting means First to first to n-1th planes passing through the midpoints of the 1st to n-1th are respectively set,
The state estimation unit is configured to determine whether the measurement value plotted by the measurement value plot unit is in a positional relationship with the kth surface (1 ≦ k <n−1) closest to the measurement value. The surface setting means sets the measured values plotted by the measured value plotting means when the driver estimates that the driver is in the kth level state. 3. The driver state estimation device according to claim 1, wherein the driver state is estimated to be in the state of the (k + 1) -th level when positioned on the sample value side of the (k + 1) -th level of the k-th surface.   The said surface setting means makes the said straight line set by the said straight line setting means a normal line, and sets the plane which passes along the midpoint of the said straight line set by the said midpoint setting means. Driver state estimation device according to.   The surface setting means uses the straight line set by the straight line setting means as a normal, passes through the midpoint of the straight line set by the midpoint setting means, and between the sample value groups having different driver states. The driver state estimation apparatus according to any one of claims 1 to 3, wherein a curved surface that passes through is set. Sample values of each of at least three parameters affected by the driver state in each of the case where the driver state is the first level and the case where the driver state is a second level better than the first level. A sample value group acquisition step of acquiring a group;
A sample value group plotting step of plotting the sample value group of each of the parameters acquired in the sample value group acquiring step on a vector space having each of the parameters as a vector axis;
On the vector space where the sample value group of the parameter is plotted by the sample value group plotting step, a first average value point that is an average value of each of the parameters in the sample value group of the first level; An average value point plotting step of plotting a second average value point that is an average value of each of the parameters in the sample value group of the second level;
A straight line setting step for setting a straight line connecting the first average value point and the second average value point plotted in the average value point plotting step;
A midpoint setting step for setting a midpoint of the straight line set in the straight line setting step;
A surface setting step for setting a plane that passes through the midpoint of the straight line set by the midpoint setting step, with the straight line set in the straight line setting step as a normal line,
A measurement value acquisition step of acquiring a measurement value of each of the parameters during operation of the driver;
A measured value plotting step of plotting the measured values of each of the parameters acquired in the measured value acquiring step on the vector space;
When the measured values plotted in the measured value plotting step are positioned on the sample value side of the first level of the surface set in the surface setting step, the driver is in the first level state. And when the measured value plotted in the measured value plotting step is located on the sample value side of the second level of the surface set in the surface setting step, the driver is in the second level. A state estimation step for estimating the state of
A driver state estimation method including: If the parameter was the vehicle and the relative velocity ΔV of the preceding vehicle, the target following distance and the inter-vehicle distance error ΔR and the accelerator pedal displacement P _a is the difference between the present inter-vehicle distance,
T _N (dP _a / dt) + P _a = H _V ΔV + H _R ΔR
In pedal delay time constant T _N in the accelerator pedal operation model _shown, the relative speed comprises a feedback gain H _V and the inter-vehicle distance error feedback gain H _R, the driver condition estimating method according to claim 6. In the sample value group obtaining step, the driver state is the first level and the driver state is the first level to the nth level (3 <n) in which the driver state is improved stepwise from the first level. A sample value group for each of the parameters in the case,
In the average value point plotting step, the first average value point to the n-th average value that are the average values of the parameters in the sample value groups of the first level to the n-th level are plotted,
In the straight line setting step, the first straight line connecting the first average value point and the second average value point plotted in the average value point plotting step to the n-1th average value point and the nth average value. Set each of the (n-1) th straight lines connecting the points,
In the midpoint setting step, the first midpoint to the n-1th midpoint of each of the first to n-1 straight lines set in the straight line setting step are set.
In the surface setting step, the first to n-1 straight lines set in the straight line setting step are normals, and the first to n-1 straight lines set in the midpoint setting step are First to first to n-1th planes passing through the midpoints of the 1st to n-1th are respectively set,
In the state estimation step, the measurement value plotted in the measurement value plotting step is the k-th surface in a positional relationship with the k-th surface (1 ≦ k <n−1) closest to the measurement value. The driver is estimated to be in the state of the k-th level, and the measured values plotted in the measured value plotting step are set in the surface setting step The driver state estimation method according to claim 6 or 7, wherein the driver is estimated to be in the state of the (k + 1) th level when positioned on the sample value side of the (k + 1) th level of the kth surface.   9. The method according to claim 6, wherein, in the surface setting step, the straight line set in the straight line setting step is set as a normal line, and a plane passing through the midpoint of the straight line set in the midpoint setting step is set. The driver state estimation method described in 1.   In the surface setting step, the straight line set in the straight line setting step is a normal line, passes through the midpoint of the straight line set in the midpoint setting step, and between the sample value groups having different driver states. The driver state estimation method according to any one of claims 6 to 8, wherein a curved surface passing through is set.

Images