JP2014102538A - Driver condition estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a driver condition estimation device capable of reducing the load required for information processing for sequentially creating a driver model and capable of increasing the speed of entire arousal level estimation processing while ensuring estimation accuracy of the arousal level.SOLUTION: An arousal level estimation system 11 according to the present invention includes a reference model setting unit 35 that sets a reference model which is a reference driver model suitable for repetitive use in association with each of plural predetermined travel scenes. A reference model operation amount acquisition unit 37 acquires a reference model steering amount by reading out a reference model corresponding to the present travel scene of a vehicle from the reference model setting unit 35 during the travel of the vehicle Ca and using the same. An arousal level estimation unit 41 estimates the arousal level of the driver based on a reference model error calculated from a difference between an actual steering angle and a reference model steering angle. With this, estimation accuracy of the arousal level can be ensured even when the travel scene of the vehicle Ca momentarily changes.

Description

  The present invention relates to a driver state estimation device for estimating the state of a driver driving a vehicle such as an automobile.

  2. Description of the Related Art Conventionally, a wakefulness estimation device has been developed for estimating the wakefulness of a driver who is driving a vehicle for the purpose of appropriately operating a vehicle such as an automobile.

  The present applicant has proposed a technique for estimating the driver's arousal level using the concept of a driver model that models driver recognition, judgment, and operation (see Patent Document 1).

  In the arousal level estimation technique according to Patent Document 1, first, the driver model creation unit uses the difference between the target azimuth angle and the actual azimuth angle (azimuth angle deviation) as a virtual driver input, and uses the actual steering angle as a virtual. As a typical driver output, a driver model representing a virtual driver input / output relationship is created. The driver model operation amount acquisition unit acquires the driver model steering angle by inputting the current azimuth deviation with respect to the driver model. The arousal level estimation unit estimates the driver's arousal level based on the difference between the actual steering angle and the driver model steering angle.

  The difference between the actual steering angle and the driver model steering angle is effective for evaluating whether the operation based on the judgment of the driver (actual steering angle) is based on the linear model (driver model steering angle). It becomes an indicator. Further, it is known that when the operation based on the judgment of the driver (actual steering angle) deviates from the driver model steering angle, the driver's arousal level is highly likely to decrease. Therefore, according to the arousal level estimation technique according to Patent Document 1, it is possible to estimate the driver's arousal level based on the difference between the actual steering angle and the driver model steering angle.

WO2011 / 040390 Publication

  However, in the arousal level estimation technique according to Patent Document 1, a driver model is sequentially generated for each predetermined period, and a residual that is basic data for arousal level estimation is calculated using the generated driver model, and obtained. A series-linked information processing process is adopted in which the driver's arousal level is estimated based on the obtained residual. For this reason, there is a problem that the load related to information processing is excessive, and the speed of the estimation process as a whole cannot be increased due to a request to perform the downstream process after waiting for the upstream process.

  The present invention has been made in view of the above circumstances, and provides a driver state estimation device that can reduce the load related to information processing related to the sequential creation of a driver model and can realize a high speed as an entire estimation process. For the purpose.

  In order to achieve the above object, an invention according to claim 1 is a driver state estimating device for estimating a state of a driver who drives a vehicle, and an operation target for acquiring an operation target value of the driver. A value acquisition unit; an actual operation amount acquisition unit that acquires an actual operation amount of the vehicle; an actual operation amount acquisition unit that acquires an actual operation amount of the driver; and an actual operation amount of the vehicle or the driver A traveling scene identification unit for identifying a current traveling scene of the vehicle from a plurality of traveling scenes that are predetermined as a traveling environment including the condition of the traveling path of the vehicle based on at least one of the actual operation amount A driver model in which a difference between the operation target value and the actual operation amount is input to the driver, the actual operation amount is output to the driver, and an input / output relationship of the driver is defined, and is used repeatedly. A suitable normative driver model A reference model setting unit that sets as a reference model, a reference model operation amount acquisition unit that acquires a reference model operation amount by inputting a difference between the operation target value and the actual operation amount to the reference model, and the actual operation amount And a driver state estimating unit that estimates the state of the driver based on a reference model error obtained from a difference in the reference model operation amount, and the reference model setting unit includes a plurality of driving scenes. The reference model is set in association with each other, and the reference model operation amount acquisition unit sets the reference model corresponding to the current traveling scene of the vehicle identified by the traveling scene identification unit to the reference model setting unit. Obtaining the reference model operation amount by inputting a difference between the operation target value and the actual movement amount with respect to the acquired reference model. The most important feature.

  The invention according to claim 2 further includes a driver model creation unit that creates the driver model, and the reference model setting unit is a driver model created in the past by the driver model creation unit. From the above, a driver model serving as a norm suitable for repeated use is set as the normative model.

  The invention according to claim 3 further includes a driver model operation amount acquisition unit that acquires a driver model operation amount by inputting a difference between the operation target value and the actual operation amount to the driver model. The reference model setting unit includes the actual operation amount and the driver model operation amount in an initial setting period until a predetermined time elapses from the start of driving of the vehicle or a change of the travel scene. The driver model used when the smallest value is obtained as the residual obtained from the difference in the driver model operation amount acquired using the acquisition unit is set as a reference model corresponding to the travel scene. It is characterized by.

  In the invention according to claim 4, the traveling scene identification unit smoothes the time series data over time series data related to at least one of the actual operation amount of the vehicle and the actual operation amount of the driver. Process and identify the current driving scene of the vehicle based on the actual movement amount data of the vehicle or the actual operation amount data of the driver and the attribute information of the plurality of driving scenes after the smoothing process. It is characterized by.

  In the invention according to claim 5, the traveling scene identification unit performs time-sequential smoothing processing on the time-series data related to the actual movement amount of the vehicle detected by the yaw rate sensor of the vehicle, A current running scene of the vehicle is identified based on the actual movement amount data after the smoothing process and the attribute information of the plurality of running scenes.

  In the invention according to claim 6, the travel scene identification unit identifies a current travel scene of the vehicle based on the actual operation amount of the driver and attribute information of the plurality of travel scenes. It is characterized by that.

  According to a seventh aspect of the present invention, the operation target value acquisition unit captures images of the driver based on road information relating to the traveling direction of the vehicle obtained by an imaging unit or obtained via a navigation device. An operation target value is acquired.

  According to the driver state estimation device according to claim 1, the effect of reducing the load related to information processing related to the sequential generation of the driver model and the speeding up as a whole of the arousal level estimation process, and the estimation accuracy related to the arousal level The effect of suppressing the decrease can be expected.

  Further, according to the driver state estimation device according to claim 2, the load reduction related to the information processing related to the sequential generation of the driver model, the effect related to the speed increase as a whole of the arousal level estimation process, and the estimation related to the arousal level In addition to the effect of suppressing the decrease in accuracy, the normative model can be easily set. In addition, if the driver model estimation device 11 is used when the driver model and the driver are the same, it is possible to expect an improvement in accuracy related to the arousal level estimation.

  In addition, according to the driver state estimation device according to claim 3, in addition to the effects of reducing the load related to information processing and increasing the speed of the overall arousal level estimation process, the reduction in estimation accuracy related to the arousal level is a high level. Can be suppressed.

  Moreover, according to the driver state estimation apparatus according to claims 4 to 6, the current traveling scene of the vehicle Ca can be accurately identified. Therefore, in addition to the effect of reducing the load related to information processing and increasing the speed of the entire arousal level estimation process, it is possible to expect the effect of suppressing the decrease in the estimation accuracy related to the arousal level at a higher level.

  In addition, according to the driver state estimation device according to claim 7, since the acquisition route related to the operation target value of the driver is clearly defined, it is possible to reduce the load related to information processing and increase the speed of the overall arousal level estimation process. In addition to the effect and the effect of suppressing the decrease in the estimation accuracy related to the arousal level, the acquisition related to the target azimuth angle (operation target value) of the driver can be accurately performed.

It is a conceptual diagram with which it uses for description of the principle which concerns on the arousal level estimation technique used as the premise of this invention. It is a conceptual diagram with which it uses for description of the principle which concerns on the arousal level estimation technique used as the premise of this invention. It is a figure where it uses for operation | movement description of the arousal level estimation technique used as the premise of this invention. It is a block diagram showing the schematic structure of the arousal level estimation apparatus (driver state estimation apparatus) which concerns on embodiment of this invention. It is a figure where it uses for operation | movement description of the arousal level estimation apparatus (driver state estimation apparatus) which concerns on embodiment of this invention. It is a flowchart figure showing the flow of a normative model setting process among operation | movement of the arousal level estimation apparatus (driver state estimation apparatus) which concerns on embodiment of this invention. It is a flowchart figure showing the flow of a wakefulness estimation process among operation | movement of the wakefulness estimation apparatus (driver state estimation apparatus) which concerns on embodiment of this invention. It is explanatory drawing showing the correspondence of the absolute value of a smoothing yaw rate, and the attribute information of a driving scene. It is explanatory drawing showing the correspondence of the attribute information of a driving | running | working scene, and the corresponding normative model. It is explanatory drawing which matches and represents the relationship of the arousal level with respect to a normative model error.

Hereinafter, a driver state estimation device according to an embodiment of the present invention will be described in detail with reference to the drawings.
It should be noted that the driver state serving as the estimation control in the driver state estimating device according to the present invention is the level related to the arousal level indicating the level related to the driver's arousal and the level of attention / concentration exerted by the driver. This is a concept that comprehensively includes the state relating to the driver. As will be described later, the arousal level in the driver state can be estimated based on the behavior of the vehicle at the time when the driver is asleep or drunken driving, for example.
Therefore, an embodiment of the driver state estimation device according to the present invention will be described by exemplifying a wakefulness estimation device that estimates the driver's wakefulness using the concept of a normative model to be described later.

[Description of Principles Related to Awakening Level Estimation Technology as a Premise of the Present Invention]
First, the principle relating to the arousal level estimation technique which is the premise of the present invention will be described with reference to FIG. 1 and FIG. FIG. 1 and FIG. 2 are conceptual diagrams for explaining the principle of the arousal level estimation technique which is a premise of the present invention. FIG. 3 is a diagram for explaining the operation of the arousal level estimation technique which is a premise of the present invention.
Note that the arousal level is one mode of the driver state as described above. For this reason, the “wakefulness estimation technique” described below can be directly applied to a driver state estimation application by replacing it with “driver state estimation technique”.

  In the example of FIG. 1, for example, it is assumed that the driver controls the azimuth of the vehicle by operating the steering wheel so that the vehicle travels along a white line drawn in the center of the road.

  First, as shown in FIGS. 1 and 2, the driver visually recognizes the azimuth deviation, which is the deviation between the target azimuth for moving the vehicle along the white line and the actual azimuth of the vehicle. Recognize through. In order to eliminate this azimuth deviation, the driver empirically determines how much and in what direction the steering wheel should be operated, and operates the steering wheel according to the determination. Then, an actual steering angle is generated in the steering wheel, and a change in azimuth occurs when the vehicle responds to the same operation. The behavior of the vehicle at this time is influenced by environmental factors (disturbances) such as the friction coefficient and inclination of the traveling road surface, the traveling speed of the vehicle, and the load capacity. A certain actual azimuth angle is generated in the vehicle.

  At this time, the actual azimuth is often different from the previous actual azimuth. Similarly, the target azimuth is often different from the immediately preceding target azimuth. Therefore, as described above, the driver recognizes and judges the azimuth deviation, which is a deviation between the target azimuth and the actual azimuth of the vehicle, from the beginning, and operates the steering wheel according to the judgment. Thereafter, the same steps as described above are repeated.

  For example, when the driver performs a drowsy driving or a drunk driving, an error occurs in the driver's recognition / judgment / operation, and it becomes difficult to accurately drive the vehicle along the white line of the road. As a result, the azimuth deviation tends to increase. Therefore, the driver's arousal level (driver state, the same applies hereinafter) can be estimated by observing the temporal transition of the azimuth deviation (control result).

  However, in the arousal level estimation method described above, the driver's arousal level is estimated without considering the environmental element (disturbance). Therefore, the estimation result inevitably includes errors due to the environmental element (disturbance). Will be included.

  Therefore, the present inventors estimate the driver's arousal level using the concept of a driver model (see FIG. 1) in which the driver's recognition / judgment / operation is modeled in consideration of environmental factors (disturbances). I devised the technology to do.

  In the arousal level estimation technique according to the present invention (awakening level estimation technique as a premise of the present invention), first, the difference between the target azimuth angle and the actual azimuth angle (azimuth angle deviation) is input to the virtual driver, A driver model is defined in which the input / output relationship of the virtual driver is defined using the steering angle as the output of the virtual driver. Since the driver model has a one-to-one input / output relationship, it can be handled as a function. When the current azimuth deviation is input to the driver model thus created, the driver model steering angle is obtained as the output of the driver model.

By the way, as described above, the difference (residual) between the actual steering angle and the driver model steering angle is based on the linear model (driver model steering angle). It becomes an effective index for evaluating whether or not. Further, it is known that when the operation based on the judgment of the driver (actual steering angle) deviates from the driver model steering angle, the driver's arousal level is highly likely to decrease.
Therefore, according to the arousal level estimation technique which is the premise of the present invention, the driver's arousal level can be estimated based on the difference (residual) between the actual steering angle and the driver model steering angle.

However, in the arousal level estimation technique that is the premise of the present invention, a driver model is sequentially created for each predetermined period, and a residual that is basic data for arousal level estimation is calculated using the created driver model, A series-linked information processing process is used in which the driver's arousal level is estimated based on the obtained residual. Specifically, as shown in FIG. 3, the driver model steering angle acquired using the driver model is used in addition to the actual steering angle for calculating the residual. That is, the residual calculation needs to be performed after the driver model creation process is completed. In other words, the driver model creation process has become a bottleneck in the arousal level estimation process.
For this reason, in the arousal level estimation technology that is the premise of the present invention, the load related to information processing is excessive, and the speed of the entire arousal level estimation process is increased from the request to perform the downstream process after waiting for the upstream process. There was a problem that could not be achieved.

[Outline of Awakening Level Estimation Device According to an Embodiment of the Present Invention]
According to the studies by the present inventors, the function required for the driver model is a function of outputting the driver model steering angle (corresponding to the “driver model operation amount” of the present invention) when the azimuth deviation is input. Even when a certain driver model is used repeatedly, it has been found that if the driver model is set appropriately, the estimation accuracy related to the arousal level is not significantly impaired.

  Therefore, in the arousal level estimation device according to the embodiment of the present invention, a driver model that serves as a norm suitable for repeated use is set as a norm model, and this norm model is repeatedly used to sequentially generate driver models. In addition to reducing the load related to information processing, the overall estimation processing speed was increased.

  Further, according to the study by the present inventors, it is known that the input / output relationship of the driver in the driver model varies depending on the traveling scene of the vehicle. Here, the traveling scene of the vehicle is an environment where the traveling vehicle is placed. For example, the situation of the traveling path of the vehicle (for example, in the case of a straight road, a curved road, or a curved road, is it gentle? , Steep, etc.), road conditions (for example, whether it is a paved road, whether it is dry, wet, frozen), weather (for example, clear weather, rain, snow, strong wind) No, the outside temperature is several times, etc.) and the concept includes whether the road is an expressway or a general road.

  A specific example of the fact that the input / output relationship of the driver in the driver model varies depending on the driving scene of the vehicle. The driver model steering angle when a common azimuth angle deviation is input, When compared with a simple curve road, the driver model steering angle on a sharp curve road tends to be larger than the driver model steering angle on a straight road.

  For this reason, for example, in a situation where the driving scene of the vehicle changes from moment to moment, if a common normative model is repeatedly used, the current driving scene of the vehicle and the characteristics of the normative model (what kind of driving scene There is a concern that the accuracy of the arousal level may be reduced.

  Therefore, in the arousal level estimation device according to the embodiment of the present invention, in association with each of a plurality of predetermined driving scenes, each set a norm model that is a driver model serving as a norm suitable for repeated use, By reading and using the reference model corresponding to the current driving scene of the vehicle while the vehicle is driving, even if the driving scene of the vehicle changes from moment to moment, the reduction in the estimation accuracy related to the arousal level is suppressed. I decided to plan.

[Block Configuration of Awakening Level Estimation Apparatus 11 According to Embodiment of the Present Invention]
Next, the block configuration of the arousal level estimation apparatus 11 according to the embodiment of the present invention will be described with reference to FIGS. 4 and 5. FIG. 4 is a block diagram showing a schematic configuration of the arousal level estimation apparatus 11 according to the embodiment of the present invention. FIG. 5 is a diagram for explaining the operation of the awakening level estimation device 11.

As shown in FIG. 4, the arousal level estimation apparatus 11 according to the embodiment of the present invention includes an imaging unit 13, a yaw rate sensor 15, a steering angle sensor 17, a speaker 19, and an ECU (electronic control unit) 20. Configured. These members constituting the awakening level estimation device 11 are mounted on the vehicle Ca.
The arousal level estimation device 11 according to the embodiment of the present invention corresponds to the “driver state estimation device” of the present invention.

  The imaging unit 13 has a function of imaging a road surface including a white line ahead of the traveling direction of the vehicle Ca. The imaging unit 13 is, for example, a CCD (individual imaging device) sensor. The road surface image information including the white line imaged by the imaging unit 13 is sent to a target azimuth angle obtaining unit 21 described later.

  The yaw rate sensor 15 has a function of detecting the yaw rate generated in the vehicle Ca (rotational angular velocity around the vertical axis of the center of gravity of the vehicle). The yaw rate detected by the yaw rate sensor 15 is sent to an actual azimuth angle acquisition unit 23 and a traveling scene identification unit 34 described later.

  The steering angle sensor 17 has a function of detecting the actual steering angle (corresponding to the “actual operation amount” of the present invention) of the steering wheel 18 operated by the driver. The steering angle sensor 17 corresponds to the “actual operation amount acquisition unit” of the present invention. The actual steering angle detected by the steering angle sensor 17 is sent to a driver model creation unit 27, a residual calculation unit 33, and a reference model error calculation unit 39, which will be described later.

  The speaker 19 generates an alarm sound according to the control operation of the alarm control unit 43 described later when the estimation result that the driver's arousal level is below a predetermined level is obtained in the awakening level estimation unit 41 described later. It has a function. This is because stimulating the auditory sense is effective in awakening a driver who is in a low arousal state.

  The ECU 20 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), an input / output circuit (including an A / D converter and a D / A converter), and the like. Is done. The CPU executes various processes including a normative model setting process and an arousal level estimation process using the RAM as a work area according to a program stored in the ROM. The flow of these processes will be described later.

  The ECU 20 includes, as logical constituent members, a target azimuth angle acquisition unit 21, an actual azimuth angle acquisition unit 23, an azimuth angle deviation calculation unit 25, a driver model creation unit 27, a bandpass filter 29, and a driver model steering angle acquisition unit. 31, a residual calculation unit 33, a running scene identification unit 34, a reference model setting unit 35, a reference model steering angle acquisition unit 37, a reference model error calculation unit 39, a wakefulness estimation unit 41, and an alarm control unit 43. Configured.

  The target azimuth angle acquisition unit 21 corresponding to the “operation target value acquisition unit” of the present invention has a function of acquiring the driver's target azimuth angle (corresponding to the “operation target value” of the present invention). Specifically, the target azimuth angle obtaining unit 21 recognizes the direction of the white line by performing image processing on the road surface image information including the white line sent from the imaging unit 13. Next, the target azimuth angle acquisition unit 21 calculates and acquires the target azimuth angle based on the amount of deviation in the yaw direction between the direction of the white line and the direction of the vehicle body longitudinal axis.

  The target azimuth angle increases as the amount of deviation in the yaw direction between the white line direction and the vehicle body longitudinal axis direction increases. The sign of the target azimuth is set so that, for example, the clockwise direction is "positive" while the counterclockwise direction is "negative", depending on the direction of the yaw direction deviation of the vehicle longitudinal axis with respect to the white line direction. Is done.

  The actual azimuth angle acquisition unit 23 corresponding to the “actual movement amount acquisition unit” of the present invention has a function of acquiring an actual azimuth angle (corresponding to the “real movement amount” of the present invention). Specifically, the actual azimuth angle acquisition unit 23 acquires the actual azimuth angle by, for example, integrating time series data of the yaw rate detected and output by the yaw rate sensor 15 over a predetermined time.

  The azimuth deviation calculation unit 25 has a function of calculating an azimuth deviation that is a difference between the target azimuth angle acquired by the target azimuth angle acquisition unit 21 and the actual azimuth angle acquired by the actual azimuth angle acquisition unit 23. The azimuth deviation calculated by the azimuth deviation calculation unit 25 is sent to the driver model creation unit 27 and the bandpass filter 29, respectively.

  As shown in FIG. 4, the driver model creation unit 27 creates a driver model that defines the input / output relationship of the driver with the azimuth angle deviation as the driver input and the actual steering angle as the driver output. It has the function to do. The driver model created in the past using the driver model creation unit 27 is necessary by adding attribute information of the driving scene to which each driver model belongs to a storage unit such as a RAM, as will be described later. Accordingly, it is stored in a readable state.

  In the present embodiment, the driver model creation unit 27 is basically configured to function only when performing the normative model setting process. In short, when the driver model creation unit 27 according to the present embodiment performs the reference model setting process for the driver model that has been sequentially created for each predetermined period in the arousal level estimation technique that is the premise of the present invention. A configuration that is created only for this is adopted. As a result, the load of information processing related to the sequential creation of the driver model is reduced, and the speed of the overall arousal level estimation process is increased.

  Incidentally, in the present embodiment, the driver model that defines the input / output relationship between the azimuth angle deviation that is the driver's input and the actual steering angle that is the driver's output is, for example, a first-order differential shown in Equation (1). Given by the equation.

[Driver model] = K / (1 + Ts) Formula (1)
However, K is a gain coefficient and Ts is a time response.

According to the definition of the driver model, the relationship of the following formula (2) is established. The solution of the first-order differential equation [K / (1 + Ts)] that is the driver model is given by the following equation (3).
[Actual steering angle] = [K / (1 + Ts)] × [azimuth angle deviation] Equation (2)
[K / (1 + Ts)] = [actual steering angle] / [azimuth angle deviation] Equation (3)

  Here, the [actual steering angle] on the right side of the equation (3) is detected and output by the steering angle sensor 17. Further, the [azimuth angle deviation] on the right side of the equation (3) is calculated by the azimuth angle deviation calculating unit 25. As described above, both values on the right side of the equation (3) can be acquired. Therefore, the driver model creation unit 27 can create a driver model by calculation by applying Equation (3).

  The band pass filter 29 selects data belonging to a frequency band of 1 to 10 rad / sec with respect to time-series data (including frequency components over a wide band) including the azimuth deviation calculated by the azimuth deviation calculator 25. It has a function to pass through automatically. Here, according to the study by the present inventors, the time series data including the azimuth deviation belonging to the frequency band of 1 to 10 rad / sec is a tendency when the driver's arousal level is lowered (for example, sleepiness is increased). Is known to represent prominently. The time-series data related to the azimuth angle deviation that has passed through the bandpass filter 29 is sent to the driver model steering angle acquisition unit 31 and the reference model steering angle acquisition unit 37, respectively.

  The driver model steering angle acquisition unit 31 corresponding to the “driver model operation amount acquisition unit” of the present invention has a function of acquiring the driver model steering angle. Specifically, the driver model steering angle acquisition unit 31 applies the above-described equation (2), and performs bandpass for the driver model [K / (1 + Ts)] created by the driver model creation unit 27. The driver model steering angle is obtained by multiplying the time-series data related to the azimuth angle deviation that has passed through the filter 29. The driver model steering angle acquired by the driver model steering angle acquisition unit 31 is sent to the residual calculation unit 33.

As shown in FIG. 3, the residual calculation unit 33 is based on the actual steering angle data sent from the steering angle sensor 17 and the driver model steering angle data acquired by the driver model steering angle acquisition unit 31. , “Residual”. Specifically, the residual calculation unit 33 obtains a difference between the actual steering angle data and the driver model steering angle data, and performs a mean square process on the obtained difference data (the obtained difference data is time-series). In the case of data, the mean square process may be performed using the time series data, and the mean square process may be performed using the obtained difference data and the accumulated past difference data. ) To calculate a “residual” as an index representing the driver's arousal level. In addition, the average value for several points on the time axis of time series data at a certain point in time, the value obtained by performing filtering processing excluding the value that exceeds the predetermined value from the center value, etc. , May be adopted as “residual”. The “residual” calculated by the residual calculation unit 33 is sent to the arousal level estimation unit 41.
In addition, the input / output which concerns on the ideal model shown in FIG. 3 means the actual input / output which a driver | operator performs. The opposite of the driver model is the actual driver as a party.

  The travel scene identification unit 34 includes a travel path condition of the vehicle Ca based on at least one of the actual azimuth angle (actual operation amount) of the vehicle Ca and the actual steering angle (actual operation amount) of the driver. From a plurality of travel scenes that are predetermined as the environment (specifically, for example, the travel path of the vehicle Ca is a straight road, a curved road, or a gentle or steep in the case of a curved road). It has a function of identifying the current running scene. Detailed operation of the traveling scene identification unit 34 will be described later.

The reference model setting unit 35 basically has a function of setting, as a reference model, a driver model serving as a reference suitable for repeated use from among the driver models previously created by the driver model creation unit 27. Have. Specifically, the normative model setting unit 35 performs the actual operation amount (actual steering angle) during the initial setting period until a predetermined time elapses from the start of driving of the vehicle Ca or when the travel scene changes. The driver used when the smallest value is obtained as the residual obtained from the difference in the driver model operation amount (driver model steering angle) acquired using the driver model operation amount acquisition unit 31. The model is set as a reference model corresponding to the travel scene. Detailed operation of the reference model setting unit 35 will be described later.
The reference model set by the reference model setting unit 35 is appropriately referred to by the traveling scene identification unit 34 and the reference model steering angle acquisition unit 37.

  The reference model set by the reference model setting unit 35 may be a driver model other than the driver model previously created by the driver model creation unit 27. “Set as a reference model” is to add attribute information of a driving scene to which each driver model belongs to a storage unit such as a RAM in preparation for repeated use of the driver model to be set. This means that it is stored in a readable state as necessary.

  The reference model steering angle acquisition unit 37 corresponding to the “reference model operation amount acquisition unit” of the present invention has a function of acquiring a reference model steering angle (corresponding to the “reference model operation amount” of the present invention). Specifically, the normative model steering angle acquisition unit 37 applies the above-described formula (2) and applies a band to the driver model [K / (1 + Ts)] set as the normative model by the normative model setting unit 35. The reference model steering angle is acquired by multiplying the time-series data related to the azimuth angle deviation that has passed through the pass filter 29. The reference model steering angle acquired by the reference model steering angle acquisition unit 37 is sent to the reference model error calculation unit 39.

As shown in FIG. 5, the reference model error calculation unit 39 is based on the actual steering angle data sent from the steering angle sensor 17 and the reference model steering angle data acquired by the reference model steering angle acquisition unit 37. It has a function of calculating a “normative model error” that serves as an index representing the driver's arousal level. Specifically, the normative model error calculation unit 39 calculates a “normative model error” by obtaining a difference between the actual steering angle data and the normative model steering angle data. The “normative model error” calculated by the normative model error calculating unit 39 is sent to the arousal level estimating unit 41.
The reference model error calculation unit 39 calculates a “reference model error” by obtaining a difference between the actual steering angle data and the reference model steering angle data and performing a mean square process on the obtained difference data. A configuration may be adopted.

The arousal level estimation unit 41 estimates the driver's arousal level based on the “residual” calculated by the residual calculation unit 33 or the “standard model error” calculated by the standard model error calculation unit 39. It has a function. Information on the arousal level estimated by the arousal level estimation unit 41 is sent to the alarm control unit 43.
The arousal level estimation unit 41 corresponds to the “driver state estimation unit” of the present invention.

  The alarm control unit 43 has a function of performing alarm control for causing the speaker 19 to generate an alarm sound when the arousal level information indicating that the driver's arousal level is below a predetermined level is sent from the arousal level estimation unit 41. Have

[Operation of Awakening Level Estimation Device 11 According to Embodiment of the Present Invention]
Next, the operation of the arousal level estimation apparatus 11 according to the embodiment of the present invention will be described with reference to FIGS. 6A, 6B and FIGS. 7A to 7C. FIG. 6A is a flowchart showing the flow of the normative model setting process among the operations of the arousal level estimation apparatus 11 according to the embodiment of the present invention. FIG. 6B is a flowchart showing the flow of the arousal level estimation process among the operations of the awakening level estimation device 11. FIG. 7A is an explanatory diagram illustrating a correspondence relationship between the absolute value of the smoothed yaw rate and the attribute information of the traveling scene. FIG. 7B is an explanatory diagram illustrating a correspondence relationship between attribute information of a traveling scene and a corresponding reference model. FIG. 7C is an explanatory diagram illustrating the relationship of the arousal level with respect to the normative model error.

  The reference model setting process shown in FIG. 6A is started from the start of driving of the vehicle Ca or the change of the running scene under the monitoring of the ECU 20. Specifically, the ECU 20 constantly monitors an on / off state of an ignition switch (not shown) and a change state of the traveling scene. The ECU 20 determines whether or not the operation of the vehicle Ca is started by checking whether or not the ignition switch has been switched from OFF to ON. Further, the ECU 20 determines whether or not the traveling scene is changed by examining the difference between the traveling scene identified by the traveling scene identification unit 34 and the immediately preceding traveling scene.

  In the example shown in FIG. 6A, data that is the basis for performing the normative model setting process is collected in an “initial setting period” (see steps S501 to S510). Here, the “initial setting period” refers to a time at which the length can be appropriately changed such as a predetermined time (for example, 10 to 30 minutes), starting from the start of driving of the vehicle Ca or when the traveling scene changes. ).

  6A and 6B have processing steps common to both (see steps S501 to S505). Therefore, in FIG. 6A and FIG. 6B, common step numbers are assigned between the processing steps common to both, and redundant description thereof is omitted.

  In step S501, the target azimuth angle acquisition unit 21 recognizes the direction of the white line by performing image processing on the image information of the road surface including the white line sent from the imaging unit 13, and recognizes the direction of the recognized white line. Based on the amount of deviation in the yaw direction from the direction of the longitudinal axis of the vehicle body, a target azimuth, which is an operation target value of the driver, is calculated and acquired.

  In step S <b> 502, the actual azimuth angle acquisition unit 23 acquires the actual azimuth angle that is the actual operation amount of the vehicle Ca by integrating the yaw rate over a predetermined time detected by the yaw rate sensor 15.

  In step S503, the azimuth deviation calculation unit 25 calculates an azimuth deviation that is the difference between the target azimuth obtained in step S501 and the actual azimuth obtained in step S502.

  In step S504, the steering angle sensor 17 detects and acquires the actual steering angle of the steering wheel 18 operated by the driver.

  In step S505, the traveling scene identification unit 34 subjects the time-series data (actual vehicle movement amount) detected by the yaw rate sensor 15 to smoothing over time, and the absolute value of the smoothed yaw rate after the smoothing process. Based on the value (actual motion amount data) and attribute information (see FIG. 7A) of a plurality of travel scenes, the current travel scene of the vehicle Ca is identified. The current travel scene of the vehicle Ca identified in step S505 is referred to when a driver model corresponding to the travel scene is created in step S506, or when a change in the travel scene is detected.

In the example of the table shown in FIG. 7A, when the absolute value of the smoothed yaw rate is 1 [deg / s] or less, “straight road” is associated as the attribute information of the traveling scene. When the absolute value of the smoothed yaw rate is in the range of 1 to 4 [deg / s], “slow curve road” is associated as attribute information of the traveling scene. When the absolute value of the smoothed yaw rate is in the range of 4 to 7 [deg / s], “steep curve road” is associated as attribute information of the traveling scene. Therefore, for example, when 3 [deg / s] is detected as the absolute value of the smoothed yaw rate, the traveling scene identification unit 34 identifies “a gentle curve road” as the current traveling scene of the vehicle Ca. .
Note that the case where the absolute value of the smoothed yaw rate exceeds 7 [deg / s] corresponds to, for example, a case where the vehicle Ca is traveling on a crank-shaped alley, or parking. In such a case, it is considered that there is almost no request for estimating the degree of arousal included in the target of the running scene. For this reason, it excludes from the object of a driving | running | working scene.

  In step S506, the driver model creation unit 27 uses the azimuth angle deviation calculated in step S503 as the driver's input, and uses the actual steering angle acquired in step S504 as the driver's output as the driving scene identified in step S505. A driver model corresponding to is created by applying Equation (3).

  In step S507, the driver model operation amount acquisition unit 31 applies the above-described equation (2) and passes the bandpass filter 29 with respect to the driver model [K / (1 + Ts)] created in step S506. The driver model steering angle is obtained by multiplying the time-series data related to the azimuth deviation.

  In step S508, the residual calculation unit 33 obtains a difference between the actual steering angle data acquired in step S504 and the driver model steering angle data acquired in step S507, and performs a mean square process on the obtained difference data. To calculate a “residual” that serves as an index representing the driver's arousal level.

In step S509, the driver model creation unit 27 stores the “residual” calculated in step S508 in association with the driver model corresponding to the travel scene identified in step S505.
Here, storing the “residual” calculated in step S508 in association with the driver model corresponding to the driving scene identified in step S505 is attribute information of the driving scene and function information of the driver model ( This means that the equation (3)) and the calculated “residual” information are stored in association with each other. A table employing such a data structure is called a driver model management table. According to the driver model management table, as will be described later, the driver model having the smallest “residual” in the initial setting period can be uniquely identified.

  In step S510, the ECU 20 determines whether or not the initial setting period has ended. This determination is performed by checking whether or not a predetermined time has elapsed from the start point of the reference model setting process (starting point of “initial setting period”) in a timer unit (not shown) placed under the monitoring of the ECU 20. .

  As a result of the determination in step S510, when it is determined that the initial setting period has not yet ended (“No” in step S510), the ECU 20 returns the process flow to step S501 and performs the following process. Let it be done. In this case (“No” in step S510), the reference model setting unit 35 does not set the reference model.

  On the other hand, if it is determined in step S510 that the initial setting period has ended ("Yes" in step S510), the ECU 20 advances the process flow to the next step S511.

In step S511, the reference model setting unit 35 refers to the driver model management table stored in the driver model creating unit 27, extracts the driver model having the smallest “residual” in the initial setting period, The extracted driver model is set as a reference model corresponding to the traveling scene identified in step S505.
Here, setting the driver model having the smallest “residual” as the reference model corresponding to the traveling scene identified in step S505 means that the attribute information of the traveling scene and the driver model having the smallest “residual” are set. Is stored in association with each other (see equation (3)). A table employing such a data structure is referred to as a normative model management table (see FIG. 7B). According to the normative model management table shown in FIG. 7B, a normative model associated with a certain traveling scene can be read out as will be described later.
When the reference model setting process in step S511 ends, the ECU 20 ends the flow of a series of processes.

  The reference model setting process described above is repeated until the setting of the reference model corresponding to each of a plurality of predetermined driving scenes is completed.

Next, the arousal level estimation process illustrated in FIG. 6B is performed, for example, when an activation switch (not illustrated) of the arousal level estimation device 11 is turned on.
Note that the processing steps of steps S501 to S505 are the same as those in FIG. 6A. For this reason, in FIG. 6B, description of the processing steps of steps S501 to S505 is omitted, and the description of the processing contents is continued from step S521.

  In step S521, the normative model setting unit 35 extracts the normative model corresponding to the traveling scene identified in step S505 using the attribute information related to the traveling scene as a search key, and the extracted normative model is illustrated in FIG. 7B. Processing to read from the management table is performed.

  In step S522, the normative model steering angle acquisition unit 37 applies the above equation (2) using the normative model read in step S521, and sets the driver model [K / By multiplying (1 + Ts)] by time-series data related to the azimuth angle deviation that has passed through the bandpass filter 29, the reference model steering angle is obtained.

  In step S523, the reference model error calculation unit 39 calculates a “reference model error” by obtaining a difference between the actual steering angle data acquired in step S504 and the reference model steering angle data acquired in step S522.

  In step S524, the arousal level estimation unit 41 estimates the driver's arousal level, which is an aspect of the driver state, based on the “normative model error” calculated by the normative model error calculation unit 39. This estimation is performed in the following procedure, for example.

  That is, first, the size of the “normative model error” is set in five levels (ER1 / ER2 / ER3 / ER4 / ER5; see FIG. 7C) in ascending order. On the other hand, the level of arousal level is changed from 5 in the order of high arousal level (for example, the order of not sleeping) (AW1 (not sleeping at all) / AW2 (somewhat sleepy) / AW3 (sleepy) / AW4 (very sleepy) / AW5 ( Very sleepy); see FIG. 7C). Based on the above setting contents, as shown in FIG. 7C, a relationship table is prepared in which ER1 is associated with AW1, ER2 is associated with AW2, ER3 is associated with AW3, ER4 is associated with AW4, and ER5 is associated with AW5.

  The awakening level estimation unit 41 identifies which stage of the ER1 to ER5 the “normative model error” calculated by the normative model error calculation unit 39 belongs to, and thus the identified “normative model error” stage. The corresponding level of alertness (assuming that it is “AW4 (very sleepy)”) is adopted as the driver's alertness.

  In step S525, the warning control unit 43 satisfies the predetermined condition (assuming that the level of arousal level is AW4 or AW5) in the level of arousal level (AW4) estimated in step S524. It is determined whether or not.

  As a result of the determination in step S525, when it is determined that the arousal level estimated in step S524 does not satisfy the predetermined condition (“No” in step S525), the alarm control unit 43 The flow is caused to jump to step S527.

  On the other hand, as a result of the determination in step S525, when it is determined that the arousal level estimated in step S524 satisfies a predetermined condition (“Yes” in step S525), the alarm control unit 43 The process flow proceeds to the next step S526. In this embodiment, the level of arousal estimated in step S524 is “AW4 (very sleepy)”, and the predetermined condition is “the level of arousal is AW4 or AW5”. Makes a determination that the arousal level estimated in step S524 satisfies a predetermined condition, and advances the process flow to the next step S526.

  In step S526, the alarm control unit 43 causes the speaker 19 to generate an alarm sound. As a result, the alarm is continuously emitted from the speaker 19 to the driver for a predetermined time (for example, 5 seconds).

  In step S526, the ECU 20 checks whether the activation switch of the arousal level estimation device 11 is on. When the activation switch of the arousal level estimation device 11 is turned on (“No” in step S527), the ECU 20 returns the process flow to step S501 and performs the following arousal level estimation process. On the other hand, when the activation switch of the arousal level estimation device 11 is turned off (“Yes” in step S527), the ECU 20 ends the processing flow.

[Operational effect of the arousal level estimation apparatus 11 according to the embodiment of the present invention]
In the arousal level estimation apparatus 11 according to the embodiment of the present invention, the target azimuth angle obtaining unit 21 describes a driver's target azimuth angle (operation target value; however, the terms in the corresponding claims are described in parentheses. The same shall apply hereinafter). The actual azimuth angle acquiring unit 23 acquires the actual azimuth angle (actual operation amount) of the vehicle Ca based on the integrated value of the yaw rate detected by the yaw rate sensor 15. The steering angle sensor 17 functioning as an actual operation amount acquisition unit of the present invention acquires the driver's actual steering angle (actual operation amount). The travel scene identification unit 34 includes a travel path condition of the vehicle Ca based on at least one of the actual azimuth angle (actual operation amount) of the vehicle Ca and the actual steering angle (actual operation amount) of the driver. The current traveling scene of the vehicle Ca is identified from among a plurality of traveling scenes that are predetermined as the environment. The reference model setting unit 35 uses a driver's input as an azimuth angle deviation that is a difference between a target azimuth angle (operation target value) and an actual azimuth angle (actual motion amount), and sets the actual steering angle (actual operation amount) as the driver's input. As an output, a driver model that defines the input / output relationship of the driver and that serves as a norm suitable for repeated use is set as a norm model.

  Here, the normative model setting unit 35 is set with a normative model that is a driver model serving as a norm suitable for repeated use in association with each of a plurality of driving scenes (see FIG. 7B). The reference model steering angle acquisition unit 37 acquires a reference model corresponding to the current traveling scene of the vehicle Ca identified by the traveling scene identification unit 34 from the reference model setting unit 35, and with respect to the acquired reference model, a target orientation is obtained. The reference model steering angle (reference model operation amount) is acquired by inputting an azimuth angle deviation which is a difference between the angle (operation target value) and the actual azimuth angle (actual movement amount). Then, the arousal level estimation unit 41 is based on a normative model error obtained from a difference between the actual steering angle (actual operation amount) and the normative model steering angle (normative model operation amount). Estimate the degree of arousal.

According to the study by the present inventors, the function required for the driver model is a function that outputs the driver model steering angle (driver model operation amount) when the azimuth deviation is input, and repeats a certain driver model. Even if it is used, it has been found that as long as the driver model is set appropriately, there is a high probability that the estimation accuracy of the arousal level is not impaired.
Therefore, the arousal level estimation apparatus 11 according to the embodiment of the present invention may adopt a configuration in which a driver model serving as a norm suitable for repeated use is set as a norm model and this norm model is used repeatedly.

The driver model set as the reference model is a driving component that is a component of the arousal level estimation device 11 mounted on the vehicle Ca on the condition that a driver model that is a reference suitable for repeated use is set as the reference model. It does not matter whether it is created by the person model creation unit 27 or so-called offline.
According to the arousal level estimation apparatus 11 according to the embodiment of the present invention, it is possible to reduce the load related to information processing related to the sequential creation of the driver model and to increase the speed of the overall arousal level estimation process.

  In addition, according to the study by the present inventors, it is known that the input / output relationship of the driver in the driver model varies depending on the traveling scene of the vehicle. For this reason, for example, in a situation where the driving scene of the vehicle changes from moment to moment, if a common normative model is repeatedly used, the current driving scene of the vehicle and the characteristics of the normative model (what kind of driving scene There is a concern that the accuracy of the arousal level may be reduced.

Therefore, in the arousal level estimation device 11 according to the present invention, a reference model that is a driver model serving as a reference suitable for repeated use is set in association with each of a plurality of predetermined driving scenes, and the vehicle You may employ | adopt the structure which reads and uses the reference | standard model corresponding to the present driving scene of a vehicle during driving | running | working.
According to the arousal level estimation apparatus 11 according to the present invention, it is possible to suppress a decrease in estimation accuracy related to the awakening level even when the traveling scene of the vehicle changes from moment to moment.

  The arousal level estimation apparatus 11 according to the embodiment of the present invention further includes a driver model creation unit 27 that creates a driver model, and the reference model setting unit 35 is created by the driver model creation unit 27 in the past. A configuration may be adopted in which a driver model serving as a norm suitable for repeated use is set as a norm model from among the driver models.

  According to the arousal level estimation apparatus 11 according to the embodiment of the present invention, in addition to the effect related to the load reduction related to information processing and the speedup of the estimation process as a whole, and the effect of suppressing the decrease in the estimation accuracy related to the arousal level. Thus, the norm model can be easily set. Moreover, if the driver model and the driver are the same, if the awakening level estimation device 11 is used, an improvement in accuracy related to the awakening level estimation can be expected.

  Further, in the arousal level estimation apparatus 11 according to the embodiment of the present invention, driving is performed by inputting an azimuth angle deviation, which is a difference between a target azimuth angle (operation target value) and an actual azimuth angle (actual motion amount), to the driver model. A driver model steering angle acquisition unit 31 that acquires the driver model steering angle (driver model operation amount) is further provided. The reference model setting unit 35 determines the actual steering angle (actual operation amount) and the driver model in an initial setting period until a predetermined time elapses from the start of driving of the vehicle Ca or when the traveling scene changes. The driver model used when the smallest value is obtained as the residual obtained from the difference in the driver model steering angle (driver model operation amount) acquired using the operation amount acquisition unit 31 You may employ | adopt the structure set as a reference | standard model corresponding to a scene.

  In the initial setting period until a predetermined time elapses from the start of driving of the vehicle Ca or when the traveling scene changes, the driver's arousal level is higher than in other periods. As a specific example, sleepiness is less likely to occur immediately after the start of driving or immediately after the change of the driving scene. In this way, the driver model that was used when the smallest residual value was obtained during the period when the driver's arousal level is relatively high (the driver model that is close to ideal with little output blurring against the input) When set as a reference model corresponding to the travel scene, it is possible to suppress a decrease in estimation accuracy related to the arousal level at a high level. This is because the driver model set as the reference model corresponding to the traveling scene is set in consideration of the fact that there is little blurring of the output with respect to the input.

  According to the arousal level estimation apparatus 11 according to the embodiment of the present invention, in addition to the effects of reducing the load related to information processing and increasing the speed of the overall arousal level estimation process, the reduction in estimation accuracy related to the arousal level is a high level. Can be suppressed.

  In the arousal level estimation apparatus 11 according to the embodiment of the present invention, the traveling scene identification unit 34 has at least the actual azimuth angle (actual operation amount) of the vehicle Ca or the actual steering angle (actual operation amount) of the driver. A time-sequential smoothing process is performed on the time series data according to any of the above, and the actual operation amount data of the vehicle Ca or the driver's actual operation amount data after the smoothing process, and attribute information of a plurality of traveling scenes ( For example, a configuration may be adopted in which the current traveling scene of the vehicle Ca is identified based on whether the traveling path of the vehicle Ca is a straight road, a gently curved road, or a sharply curved road.

  Further, the traveling scene identification unit 34 performs time-sequential smoothing processing on time-series data related to the actual azimuth angle (actual motion amount) of the vehicle Ca detected by the yaw rate sensor 15 of the vehicle Ca, and performs the smoothing. A configuration may be adopted in which the current traveling scene of the vehicle Ca is identified based on the processed actual azimuth (actual movement amount) data and attribute information of a plurality of traveling scenes.

  Further, the traveling scene identification unit 34 adopts a configuration that identifies the current traveling scene of the vehicle Ca based on the driver's actual steering angle (actual operation amount) and attribute information of a plurality of traveling scenes. May be.

  According to the arousal level estimation apparatus 11 according to the embodiment of the present invention, the current traveling scene of the vehicle Ca can be accurately identified. Therefore, in addition to the effect of reducing the load related to information processing and increasing the speed of the entire arousal level estimation process, it is possible to expect the effect of suppressing the decrease in the estimation accuracy related to the arousal level at a higher level.

  Moreover, in the arousal level estimation apparatus 11 which concerns on embodiment of this invention, the target azimuth angle acquisition part 21 is imaged by the imaging part 13, or in the advancing direction of the vehicle Ca obtained via a navigation apparatus (not shown). A configuration in which the driver's target azimuth (operation target value) is acquired based on the traveling road information may be employed.

As a navigation device, for example, it has a positioning function by autonomous navigation and a positioning function by satellite navigation using a GPS (Global Positioning System) receiver, and uses the autonomous navigation and satellite navigation to determine the current position of the host vehicle. An in-vehicle navigation device having a detection function is used. The target azimuth angle acquisition unit 21 can acquire the driver's target azimuth angle (operation target value) based on travel road information related to the traveling direction of the vehicle Ca obtained via the navigation device.
In addition, if the positioning function by satellite navigation using a DGPS (Differential Global Positioning System) receiver is used, positioning accuracy can be improved.

  According to the arousal level estimation device 11 according to the embodiment of the present invention, the acquisition route related to the operation target value of the driver is clearly defined, so that the load reduction related to information processing and the overall speed of the awakening level estimation process can be improved. In addition to the effect and the effect of suppressing the decrease in the estimation accuracy related to the arousal level, the acquisition related to the target azimuth angle (operation target value) of the driver can be accurately performed.

[Other Embodiments]
The embodiment described above shows an embodiment of the present invention. Therefore, the technical scope of the present invention should not be limitedly interpreted by these. This is because the present invention can be implemented in various forms without departing from the gist or main features thereof.

  For example, when a so-called offline created driver model is set as a reference model, the driver model creating unit 27, the driver model steering angle obtaining unit 31, and the residual calculation included in the ECU 20 of the arousal level estimation device 11 The part 33 can be omitted.

  In the description of the embodiment according to the present invention, the target azimuth is exemplified as the operation target value, but the present invention is not limited to this example. As the operation target value, a target displacement amount that is a target value of the displacement amount with respect to the horizontal direction of the white line may be used. In this case, the actual operation amount may be replaced with the actual displacement amount, and the actual operation amount may be replaced with the actual steering angle.

In the description of the embodiment according to the present invention, the ECU 20 included in the arousal level estimation device 11 includes, as logical components (software), a target azimuth angle acquisition unit 21, an actual azimuth angle acquisition unit 23, and an azimuth angle deviation calculation. Unit 25, driver model creation unit 27, bandpass filter 29, driver model steering angle acquisition unit 31, residual calculation unit 33, travel scene identification unit 34, reference model setting unit 35, reference model steering angle acquisition unit 37, Although an example having the reference model error calculation unit 39, the arousal level estimation unit 41, and the alarm control unit 43 has been described, the present invention is not limited to this example.
These various functional units 21, 23, 25, 27, 29, 31, 33, 34, 35, 37, 39, 41, and 43 may be configured by hardware.

  Further, in the description of the embodiment according to the present invention, the example of setting the arousal level to five levels has been described, but the present invention is not limited to this example. The awakening level can be set to any number of levels.

  In the description of the embodiment according to the present invention, the alarm control unit 43 has been described with an example in which the speaker 19 generates an alarm sound when the arousal level satisfies a predetermined condition. The invention is not limited to this example. The warning that promotes the driver's awakening may employ a mode in which the driver's vision, touch, or smell is stimulated in addition to hearing. Moreover, you may employ | adopt the aspect which stimulates several combinations simultaneously among a driver | operator's auditory sense, visual sense, tactile sense, or olfactory sense.

  Further, in the description of the embodiment according to the present invention, as the variation of the attribute information related to the traveling scene of the vehicle Ca, the curvature information of the traveling road (“straight road”, “gradual curved road”, “steep curved road”). However, the present invention is not limited to this example. As variations of attribute information related to the traveling scene of the vehicle Ca, in addition to the curvature information of the traveling road, the situation information on the traveling road surface (for example, whether it is a paved road, whether it is dry, wet, frozen) Or weather information (for example, whether it is fine weather, rain, snow, strong winds, the outside temperature is several times, etc.), type information regarding whether the road is a highway or a general road, etc. .

  Finally, in the description of the present embodiment, the wakefulness estimation device 11 has been described as an example for realizing the driver state estimation device according to the present invention, but the present invention is not limited to this example. . The driver state estimation device according to the present invention uses all the functions that estimate the state of a driver who is performing a nap driving, drunk driving, aside driving, random driving, etc., using the concept of a normative model. This is because it is a comprehensive concept.

11 Arousal level estimation device (driver state estimation device)
13 Imaging unit 15 Yaw rate sensor 17 Steering angle sensor (actual operation amount acquisition unit)
18 Steering wheel 19 Speaker 20 ECU
21 Target azimuth angle acquisition unit (operation target value acquisition unit)
23 Real azimuth angle acquisition unit (actual movement amount acquisition unit)
25 Azimuth Deviation Calculation Unit 27 Driver Model Creation Unit 29 Band Pass Filter 31 Driver Model Steering Angle Acquisition Unit (Driver Model Operation Amount Acquisition Unit)
33 residual calculation unit 34 traveling scene identification unit 35 reference model setting unit 37 reference model steering angle acquisition unit (reference model operation amount acquisition unit)
39 Reference model error calculation unit 41 Arousal level estimation unit (driver state estimation unit)
43 Alarm control unit

Claims (7)

A driver state estimating device for estimating a state of a driver driving a vehicle,
An operation target value acquisition unit for acquiring the operation target value of the driver;
An actual operation amount acquisition unit for acquiring an actual operation amount of the vehicle;
An actual operation amount acquisition unit for acquiring the actual operation amount of the driver;
Based on at least one of the actual operation amount of the vehicle or the actual operation amount of the driver, a plurality of traveling scenes that are predetermined as a traveling environment including the state of the traveling path of the vehicle are used. A driving scene identification unit for identifying the current driving scene;
A driver model in which the difference between the operation target value and the actual operation amount is input to the driver, the actual operation amount is output to the driver, and the input / output relationship of the driver is defined, and is suitable for repeated use. A normative model setting unit for setting a standard driver model as a normative model;
A normative model operation amount acquisition unit that acquires a reference model operation amount by inputting a difference between the operation target value and the actual motion amount to the reference model;
A driver state estimating unit that estimates the state of the driver based on a reference model error determined from a difference between the actual operation amount and the reference model operation amount;
With
In the reference model setting unit, the reference model is set in association with each of the plurality of driving scenes,
The reference model operation amount acquisition unit acquires the reference model corresponding to the current driving scene of the vehicle identified by the driving scene identification unit from the reference model setting unit, and for the acquired reference model, Obtaining the reference model operation amount by inputting a difference between the operation target value and the actual operation amount;
A driver state estimation device characterized by the above. The driver state estimating device according to claim 1,
A driver model creating unit for creating the driver model;
The normative model setting unit sets a driver model serving as a norm suitable for repeated use as the normative model from among the driver models created in the past by the driver model creating unit,
A driver state estimation device characterized by the above. The driver state estimation device according to claim 2,
A driver model operation amount acquisition unit for acquiring a driver model operation amount by inputting a difference between the operation target value and the actual operation amount to the driver model;
The reference model setting unit includes the actual operation amount and the driver model operation amount in an initial setting period until a predetermined time elapses from the start of driving of the vehicle or a change of the travel scene. The driver model used when the smallest value is obtained as the residual obtained from the difference in the driver model operation amount acquired using the acquisition unit is set as a reference model corresponding to the travel scene. To
A driver state estimation device characterized by the above. The driver state estimation device according to any one of claims 1 to 3,
The travel scene identification unit performs time-sequential smoothing processing on time series data related to at least one of the actual operation amount of the vehicle or the actual operation amount of the driver, and after the smoothing processing Based on the actual operation amount data of the vehicle or the actual operation amount data of the driver and the attribute information of the plurality of traveling scenes, the current traveling scene of the vehicle is identified.
A driver state estimation device characterized by the above. The driver state estimation device according to claim 4,
The traveling scene identification unit performs time-sequential smoothing processing on time series data related to the actual motion amount of the vehicle detected by the yaw rate sensor of the vehicle, and actual motion amount data after the smoothing processing, And, based on the attribute information of the plurality of driving scenes, to identify the current driving scene of the vehicle,
A driver state estimation device characterized by the above. The driver state estimation device according to any one of claims 1 to 3,
The traveling scene identifying unit identifies a current traveling scene of the vehicle based on the actual operation amount of the driver and attribute information of the plurality of traveling scenes.
A driver state estimation device characterized by the above. The driver state estimation device according to any one of claims 1 to 6,
The operation target value acquisition unit acquires the driver's operation target value based on traveling road information related to the traveling direction of the vehicle obtained by an imaging unit or obtained via a navigation device.
A driver state estimation device characterized by the above.

Images