JP6924629B2 - Automatic control device and its control method - Google Patents

Description

The present invention relates to an automatic control device and a control method thereof.

In recent years, the ECU (Electronic Control Unit) of a moving body such as a vehicle automatically determines the scene in which the vehicle is placed and the environmental conditions, and drives autonomously, regardless of the manual operation of the driver (individual). Development of automatic driving technology is in progress.
In this kind of automatic driving technology, the ECU of the moving body learns the individual characteristics of the driver in various scenes and performs automatic driving close to the individual characteristics of the driver so as not to give the driver a sense of discomfort. ing.

Patent Document 1 discloses an automatic control device for a moving body.

JP-A-2007-176396

In the automatic control device for a moving body disclosed in Patent Document 1, a model adapted to the individual characteristics of the driver is constructed by learning the individual characteristics (operation characteristics) of the driver in various scenes. As a result, in this automatic control device, the driver's individual characteristics are reflected in the automatic driving, so that the driver's discomfort can be reduced.

However, in the automatic control device for a moving body of Patent Document 1, there is no criterion for determining whether or not the learning of the individual characteristics of the driver is completed.
Therefore, in this automatic control device, as a result of delaying the determination of the completion of learning of the individual characteristics of the driver, the reflection of the individual characteristics in the automatic operation is delayed. Further, in this automatic control device, there is a possibility that the individual characteristics of the driver are reflected in the automatic driving in a state where the learning of the individual characteristics of the driver is insufficient, and in this case, there is a problem that the driver feels uncomfortable. There is.

Therefore, the present invention has been made in view of the above problems, and the automatic control device can quickly reflect the individual characteristics in the automatic driving by appropriately determining the completion of learning the individual characteristics of the driver for each scene. The purpose is to do. In addition, it is intended to reduce the sense of discomfort given to the driver by performing automatic driving for each scene where it is determined that the learning of the individual characteristics of the driver is completed.

In order to solve the above problems, an individual authentication unit that authenticates an individual who operates the device, a scene determination unit that determines the scene in which the device is placed, an individual characteristic learning unit that learns the individual characteristics of an individual, and an individual In a predetermined scene determined by the manual operation estimation unit that estimates the manual operation of the device by the individual based on the individual characteristics of the individual and the scene determination unit, the estimation manual estimation by the manual operation estimation unit based on the individual characteristics of the individual The learning progress determination unit that determines the learning progress in a predetermined scene by comparing the operation with the actual manual operation of the device by the individual, and the learning progress for each predetermined scene determined by the scene determination unit. Based on the determination result of the above, the automatic control device is provided with a control unit for determining the execution or non-execution of the automatic control adapted to the individual characteristics of the individual for each scene.

According to the present invention, in the automatic control device, the individual characteristics can be quickly reflected in the automatic driving by appropriately determining the completion of learning the individual characteristics of the driver for each scene. Further, by performing automatic driving for each scene where it is determined that the learning of the individual characteristics of the driver is completed, it is possible to reduce the discomfort given to the driver.

It is a block diagram explaining the automatic control device which concerns on embodiment. It is a flowchart of individual adaptation control. It is a flowchart of the update process of the past information database. It is a flowchart of the update process of the scene judgment standard database. It is a flowchart of the process of relearning of a scene judgment standard. It is a conceptual diagram explaining the method of re-learning of a scene judgment standard. It is a flowchart of an individual characteristic learning process. It is a flowchart of the judgment process of the learning completion of an individual characteristic in a learning progress degree determination part. It is a figure explaining an example which determines the transition permission to a learning completion state based on a learning progress degree. It is a block diagram explaining a safety determination part. It is a figure explaining an example of the judgment of the occurrence of a dangerous event by a dangerous event determination part. It is a block diagram explaining the scene match degree verification part. It is a block diagram explaining the automatic control apparatus which concerns on 2nd Embodiment. It is a flowchart of individual adaptation control which concerns on 2nd Embodiment. It is a flowchart of the update process of the past information database which concerns on 2nd Embodiment. It is a flowchart of the weight determination process which concerns on 2nd Embodiment.

Hereinafter, the automatic control device 1 according to the embodiment of the present invention will be described.
In the following description, a case where the automatic control device 1 is applied to automatic driving of a moving body (not shown) such as a vehicle will be described as an example.
FIG. 1 is a block diagram illustrating an automatic control device 1 according to an embodiment.

[Automatic control device]
As shown in FIG. 1, the automatic control device 1 includes a scene judgment reference database 10, a scene judgment unit 11, an individual authentication unit 12, a past information database 13, an individual characteristic learning unit 14, and a manual operation estimation unit 15. And a learning progress determination unit 16 and a control unit 17.

In the scene determination standard database 10, parameters used for classifying scenes in automatic driving are stored corresponding to the scenes to be classified. Therefore, by acquiring a predetermined parameter from the scene determination standard database, it is possible to extract the scene corresponding to this parameter.

Here, the scene means a road shape on which a moving body such as a vehicle travels, and is a concept including a certain distance or time length. For example, examples of scenes include short straight line scenes, long straight line scenes, gentle curve scenes, sharp curve scenes, uphill scenes, downhill scenes, and intersection scenes. And each scene includes not only one scene but also a certain distance or time length.

The scene judgment standard database 10 described above may be provided in a control device or the like on the moving body (vehicle) side, but the scene judgment standard database 10 may be installed in an external server (not shown) located at a location different from the moving body. It is more preferable to provide.

By providing the scene judgment reference database 10 on an external server (not shown) and connecting the external server and the ECU of the mobile body (not shown) via the information network, the ECU of the mobile body can be used as needed. Therefore, the parameters (scenes) stored in the scene determination reference database 10 in the external server can be read out.

By doing so, the memory capacity of the control device on the moving body side can be reduced, and the increase in the manufacturing cost and the maintenance / management cost of the control device on the moving body side can be suppressed. In addition, the computational resources of the control device on the moving body side can be used intensively for highly urgent processing such as danger avoidance processing. Further, even when the driver gets on another moving body (vehicle), it is possible to read necessary information from an external server via the information network and perform individual adaptation control.

A scene determination unit 11 is connected to the scene determination standard database 10, and the scene determination unit 11 includes parameters used for classifying scenes stored in the scene determination standard database 10 and environmental information 50. , Is entered.

The scene determination unit 11 determines the scene in which the moving object (vehicle) is placed, based on the parameters input from the scene determination reference database 10 and the environment information 50.

Here, the environmental information 50 includes, for example, a road shape (straight line, curve, slope, presence / absence of an intersection, etc.), density of other moving objects (other vehicles) in the vicinity, and moving speed (vehicle speed) of other moving objects in the vicinity. ), Distance to surrounding obstacles, position in the width direction of the vehicle in the lane of the road, temperature, season, brightness, and other external environmental information. The environmental information 50 also includes internal environmental information such as the vehicle speed, acceleration, steering angle, pitch, and roll of the own vehicle.

The individual authentication unit 12 acquires the individual authentication information 52 of the driver (individual), and collates the acquired individual authentication information 52 with the individual authentication information already registered in the past information database 13 described later. , Authenticate the driver.

The individual authentication information 52 is not particularly limited as long as it is information that can identify the driver, and may be, for example, identification information such as a personal ID, a personal identification number (password), and a driver's license. The individual authentication information 52 is stored in the memory (not shown) of the ECU until the ignition key and the start / stop button are turned off.

The individual authentication unit 12 is connected to the past information database 13, and the individual authentication unit 12 reads out the individual authentication information already registered in the past information database 13.

In addition to the driver's individual authentication information 52, the scene (judgment result) determined by the scene determination unit 11, the environment information 50, and the manual operation information 51 are input and stored in the past information database 13. ..

The past information database 13 specifies a memory address (memory area) corresponding to the individual authentication information 52 in the past information database 13 based on the individual authentication information 52 of the driver authenticated by the individual authentication unit 12. Then, the past information database 13 stores the scene (determination result) determined by the scene determination unit 11 in association with the specified memory address, the environment information 50 at that time, and the manual operation information 51 of the driver. ..

On the other hand, when the past information database 13 determines that the individual authentication information 52 of the driver authenticated by the individual authentication unit 12 has not been registered (unregistered) in the past, the individual authentication information 52 of the unregistered driver A new memory address (memory area) corresponding to the above is created, and the scene (judgment result) determined by the scene determination unit 11, the environment information 50 at that time, and the driver's manual operation information 51 are added to the memory address. And are associated and memorized.

Here, the manual operation information 51 includes, for example, steering angle of steering, vehicle speed, ON / OFF operation of each light and each lamp, set temperature of air conditioning, music selection and volume of in-vehicle audio, car navigation operation information, and the like. When the opening / closing amount of the window and the light-shielding property of the window can be arbitrarily set by applying a voltage, the voltage value (light-shielding degree) and the like can be given as an example.

The past information database 13 is connected to the individual characteristic learning unit 14, and each information stored in the past information database 13 is input to the individual characteristic learning unit 14.

The individual characteristic learning unit 14 is specified by the individual authentication information 52 based on the scene (determination result) stored in the past information database 13, the environment information 50, the manual operation information 51, and the individual authentication information 52. Generate an individual characteristic model of the driver.

Specifically, the individual characteristic learning unit 14 learns what kind of manual operation the driver has performed in each scene by a method such as machine learning, and generates an individual characteristic model.
For learning individual characteristics (operation characteristics), well-known methods such as machine learning and deep learning can be used.

The individual characteristic learning unit 14 is connected to the manual operation estimation unit 15, and the individual characteristic model generated by the individual characteristic learning unit 14 is input to the manual operation estimation unit 15.

In addition to the individual characteristic model generated by the individual characteristic learning unit 14, the current environmental information 50 is input to the manual operation estimation unit 15, and the driver's operation is based on the individual characteristic model and the environmental information 50. Estimate manual operation.

Here, the driver's manual operation includes, for example, steering operation, accelerator pedal operation, ON / OFF operation of each light and each lamp, air conditioning operation, in-vehicle audio operation, car navigation operation, window opening / closing operation, and the like. Is given as an example.

The manual operation estimation unit 15 is connected to the learning progress determination unit 16, and the driver's manual operation information (hereinafter, also referred to as estimated manual operation information) estimated by the manual operation estimation unit 15 is the learning progress. It is input to the determination unit 16.

In the learning progress determination unit 16, in addition to the estimated manual operation information estimated by the manual operation estimation unit 15, the manual operation information 51 by the current driver and the scene determined by the scene determination unit 11 (determination result) And are entered.

The learning progress determination unit 16 sets the difference R (estimated manual operation information-manual operation information 51) between the estimated manual operation information estimated by the manual operation estimation unit 15 and the current driver's manual operation information 51 as a scene. By calculating for each, the learning progress for each scene is determined.
For example, when the difference R of the above-mentioned comparison result is equal to or less than a predetermined threshold value Rth (R ≦ Rth), the learning progress determination unit 16 determines that the learning progress is high (learning completed), and the difference R is greater than the threshold value Rth. When is also large (R> Rth), it is determined that the learning progress is low (learning is not completed), and the determination result of the learning progress is output to the control unit 17.

The control unit 17 determines whether or not to implement the individual adaptation control for each scene based on the determination result of the learning progress degree determined by the learning progress degree determination unit 16.

When the learning progress determination unit 16 determines that the learning progress is high (difference R ≦ threshold value Rth), the control unit 17 executes individual adaptation control.

Here, an example of the execution conditions of the individual adaptation control by the control unit 17 will be described.
First, the scene determination unit 11 includes at least two or more discontinuous first scenes (for example, gentle curve scenes) and second scenes (for example, intersection scenes) included in a predetermined scene (for example, a curve scene). Is determined.

Then, the learning progress determination unit 16 determines the learning progress of the individual characteristics of the driver in the first scene (for example, a gentle curve scene) and the second scene (for example, an intersection scene).

As a result, the control unit 17 determines that both the learning progress determination result in the first scene and the learning progress determination result in the second scene are in the learning completed state (learning progress is high). When it is determined by the learning progress determination unit 16, individual adaptation control is performed.

In this way, the control unit 17 determines whether or not to implement the individual adaptive control based on whether or not the learning progress in two or more discontinuous scenes with completely different scenes is in the completed state. It can be accurately determined that the learning of a predetermined scene including two or more discontinuous scenes has been completed.

Further, the scene determination unit 11 includes a third scene (for example, a first straight line scene) and a fourth scene (for example, a second straight line scene) included in a predetermined scene (for example, a straight line scene) at positions separated from each other. ).

Then, the learning progress determination unit 16 determines the learning progress of the individual characteristics of the driver in the third scene (for example, the first straight line scene) and the fourth scene (for example, the second straight line scene). ..

As a result, the control unit 17 determines that both the learning progress determination result in the third scene and the learning progress determination result in the fourth scene are in the learning completed state (learning progress is high). When it is determined by the learning progress determination unit 16, individual adaptation control is performed.

In this way, the control unit 17 determines whether or not to implement the individual adaptation control based on whether or not the learning progress in the two or more scenes at a distance is completed. It can be accurately determined that the learning of a predetermined scene including the discontinuous scenes at a distance of is completed.

The learning progress determination unit 16 determines the learning progress in two scenes (fifth scene, sixth scene) separated in time, and the control unit 17 determines the learning progress in two scenes (fifth scene, fifth scene). Individual adaptation control may be performed when the learning progress in (6 scenes) is in the learning completion state.

On the other hand, when it is determined that the learning progress is low (difference R> threshold value Rth), the control unit 17 does not perform individual adaptation control, but executes automatic operation control that is initially set (default) in advance. ..

Here, the individual adaptive control is an automatic driving control adapted to the individual characteristics (operation characteristics) of the driver based on the individual characteristics model generated by learning the individual characteristics (operation characteristics) of the driver (individual). Means to do.

The control unit 17 calculates the learning efficiency K from the learning progress, and when the learning efficiency K is equal to or higher than a predetermined threshold value Kth (K ≧ Kth), determines that the learning progress is high, and the learning efficiency K is a predetermined threshold value. If it is smaller than Kth (K <Kth), it may be determined that the learning progress is low.

The above-mentioned past information database 13, individual characteristic learning unit 14, manual operation estimation unit 15, and learning progress determination unit 16 may be provided in a control device on the moving body (vehicle) side, but these are used as moving bodies. It is provided on an external server (not shown) located at a different location from the above, and the external server and the ECU of the mobile body (not shown) are connected via an information network, and if necessary, the ECU of the mobile body can be used. It is preferable to read out this information.

By doing so, the memory capacity of the control device on the moving body side can be reduced, and the increase in the manufacturing cost and the maintenance / management cost of the control device on the moving body side can be suppressed. In addition, the computational resources of the control device on the moving body side can be used intensively for highly urgent processing such as danger avoidance processing. Further, even when the driver gets on another moving body (vehicle), it is possible to read necessary information from an external server via the information network and perform individual adaptation control.

[Individual adaptation control]
Next, the processing of individual adaptation control in the automatic control device 1 will be described.
FIG. 2 is a flowchart of individual adaptation control.

First, in step S101, the ECU (not shown) of the moving body determines whether or not the automatic operation of the moving body is permitted. When the ECU determines that the automatic operation is permitted (step S101: Yes), the ECU proceeds to step S102.
Here, although various methods can be considered, the permission for automatic driving is set by the driver turning on the switch for setting the automatic driving mode.

In step S102, the mobile ECU determines whether or not individual adaptive control is permitted. When the ECU determines that the individual adaptive control is permitted (step S102: Yes), the ECU proceeds to step S103.

If the mobile ECU determines in step S102 that individual adaptation control is not permitted (step S102: No), the process proceeds to step S106. In step S106, the automatic control device 1 executes the automatic operation of the moving body by the automatic operation control that is initially set (default) in advance.

In step S103, the learning progress determination unit 16 determines in the currently traveling scene based on the scene (determination result) determined by the scene determination unit, the environment information 50, the manual operation information 51, and the individual authentication information 52. It is determined whether or not the learning of the individual characteristics of the driver is completed.

In step S103, when the learning progress determination unit 16 determines that the learning of the individual characteristics of the driver is completed (step S103: Yes), the learning progress determination unit 16 proceeds to step S104. In step S104, the control unit 17 executes individual adaptation control according to the individual characteristics of the driver in the scene based on the determination of the learning completion by the learning progress determination unit 16.

In step S103, when the learning progress determination unit 16 determines that the learning of the individual characteristics of the driver has not been completed (step S103: No), the learning progress determination unit 16 proceeds to step S106. In step S106, the control unit 17 does not perform the individual adaptation control based on the determination result of the learning incompleteness by the learning progress determination unit 16, but the initial setting (default) of the automatic operation control of the moving body. Carry out automatic operation.

In step S101 described above, the ECU (not shown) of the moving body sets the switch for manually setting the automatic operation mode to OFF by the driver, or the moving body automatically operates according to the environmental information 50. It is set to OFF when it is determined that the continuation cannot be continued. In this case, the ECU of the moving body determines that automatic operation is not permitted (step S101: NO), and proceeds to step S105.

In step S105, the mobile ECU acquires the current environment information 50 and the current driver's manual operation information 51, transmits the information to the past information database 13, and ends the process.

[Update past information database]
Next, the update process of the past information database 13 will be described.
FIG. 3 is a flowchart of the update process of the past information database 13.

In step S201, the individual authentication unit 12 acquires the individual authentication information 52 of the current driver.

In step S202, the individual authentication unit 12 determines whether or not the individual authentication information 52 of the current driver acquired in step S201 is already stored in the past information database 13. When the individual authentication unit 12 determines that the individual authentication information 52 is already stored in the past information database 13 (step S202: Yes), the individual authentication unit 12 proceeds to step S203.

In step S203, the individual authentication unit 12 acquires the memory address on the past information database 13 in which the individual authentication information 52 of the driver is stored, and proceeds to step S204.

On the other hand, when the individual authentication unit 12 determines that the individual authentication information 52 is not stored in the past information database 13 (step S202: No), the individual authentication unit 12 proceeds to step S209.

In step S209, the individual authentication unit 12 sets a new memory address for storing the individual authentication information 52 of the current driver on the past information database 13, and proceeds to step S204.

In step S204, the scene determination unit 11 acquires the environment information 50 and proceeds to step S205.

In step S205, the scene determination unit 11 determines the scene based on the parameters necessary for determining the scene acquired from the scene determination reference database 10 and the environment information 50 acquired in step S204.

In step S206, the ECU (not shown) of the moving body determines whether or not the moving body is undergoing manual operation (manual operation) by the driver. When the ECU of the moving body determines that the manual operation is in progress (step S206: Yes), the process proceeds to step S207, and in step S207, the current operation information (manual operation information 51) by the driver's manual operation is acquired.

Then, in step S208, the moving unit ECU (not shown) acquires the acquired environment information 50, the manual operation information 51, and the scene (determination result) determined by the scene determination unit 11 in steps S203 and S209. It is stored in the memory address on the past information database 13 and the process is terminated.

As a result, when the driver manually operates, the environment information 50 and the manual operation information 51 stored in the past information database 13 and the scene (judgment result) determined by the scene determination unit 11 are acquired. Each of the acquired information is stored in the past information database 13. Therefore, in the past information database 13, each of these pieces of information is accumulated (updated) without the driver feeling annoyed.

If the ECU of the moving body (not shown) determines in step S206 that the moving body is not in manual operation (automatic operation is in progress) (step S206: No), the process proceeds to step S210 to obtain environmental information. The scene (determination result) determined by the 50 and the scene determination unit 11 is transmitted to the control unit 17 to end the process.

[Scene judgment criteria database update process]
Next, the update process of the scene determination standard database 10 will be described.
In the automatic control device 1, when there is a scene in which the learning efficiency K of the individual characteristics of the driver is extremely low, the scene stored in the scene judgment reference database 10 is changed to another scene, and the scene judgment reference database 10 is updated. I do.
FIG. 4 is a flowchart of the update process of the scene determination reference database 10.

In step S301, the learning progress determination unit 16 calculates the learning efficiency K for each scene for all the drivers (individuals) stored in the past information database 13.

Here, the learning progress determination unit 16 sets a smaller value when the learning progress T is sufficiently long but the learning progress is low, or when the learning time T is short but the improvement rate of the learning progress is low. Calculated as learning efficiency K.

In step S302, in any driver (individual), the learning progress determination unit 16 has a learning time T of a predetermined threshold value Tth or more (T ≧ Tth) and a learning efficiency K of a predetermined threshold value Kth. It is determined whether or not there is a scene where the value is less than (K <Kth).

When the learning progress determination unit 16 determines that there is a scene satisfying the condition of step S302 (step S302: Yes), the learning progress determination unit 16 proceeds to step S303. On the other hand, when the learning progress determination unit 16 determines that there is no scene satisfying the above-mentioned conditions (step S302: No), it determines that there is no scene with extremely low learning efficiency K, and ends this process.

In step S303, the learning progress determination unit 16 has a driver having a scene in which the learning time T is equal to or greater than the predetermined threshold value Tth and the learning efficiency K is less than the predetermined threshold value Kth in the same predetermined scene. It is determined whether or not the number N of (individuals) is a majority (N ≧ M / 2) of the number M of all the drivers (individuals) stored in the past information database 13.

In step S303, when the learning progress determination unit 16 determines that the number N of individuals satisfying the above-mentioned conditions is a majority or more of the whole (step S303: Yes), the process proceeds to step S304.

In step S304, the moving unit ECU (not shown) relearns the scene determination criteria based on the environmental information 50 and the manual operation information 51 of all the individuals. The re-learning process (method) of this scene determination criterion will be described later.

Then, in step S305, the moving ECU updates the scene of the scene determination reference database 10 (changes to another scene, deletes the scene, etc.) based on the result of relearning in step S304, and processes the scene. To finish.
As a result, the automatic control device 1 can update the scenes in the scene determination reference database 10 to more appropriate ones by reviewing or deleting the scenes having a remarkably low learning efficiency K.

In step S303, there is a scene in which the number N of drivers (individuals) whose learning efficiency K is less than the predetermined threshold value Kth is not a majority of the number M of all drivers (individuals) (step S303). : No), the scene is determined by the driver (individual) based on the environmental information 50 and the manual operation information 51 collected by the driver (individual) having the scene having the learning efficiency K remarkably low in step S306. Relearn the criteria.
Here, the re-learning of the scene determination criteria in the individual driver (individual) is performed by the same process (method) as the re-learning of the scene determination criteria in step S304.

Then, the process proceeds to step S305, and the moving unit ECU (not shown) updates the individual scenes in the scene judgment criteria database 10 based on the result of re-learning of the scene judgment criteria in step S306, and ends the process. do.

[Re-learning of scene judgment criteria]
Next, the process of re-learning the scene determination criteria described in step S304 (step S306) will be described.
FIG. 5 is a flowchart of the process of re-learning the scene determination criteria.
FIG. 6 is a conceptual diagram illustrating a method of re-learning the scene determination criteria.

As shown in FIG. 5, in step S401, the mobile ECU (not shown) applies an individual characteristic model applied in a predetermined scene (past scene) determined to have a significantly low learning efficiency K to the external network. Temporarily save to the connected memory (not shown).

In step S402, the moving ECU (not shown) reclassifies a predetermined scene with a significantly low learning efficiency K.

Specifically, as shown in FIG. 6, when the moving body ECU determines, for example, in the curve scene A and the intersection scene C that the learning efficiency K of the driver is extremely low, the curve scene A is gently changed. The curve scene A1 and the sharp curve scene A2 are reclassified, and the intersection scene C is reclassified into the three-way intersection scene C1 and the five-way intersection scene C2.
In the embodiment, since the learning progress determination unit 16 determines that the learning efficiency K of the straight line scene B is high, the mobile ECU does not reclassify the straight line scene B.

Returning to FIG. 5, in step S403, the learning progress determination unit 16 reads out the individual characteristic model saved in the memory (not shown) in step S401 and applies it to the reclassified scene.

In step S404, the learning progress determination unit 16 determines whether or not the learning efficiency K based on the individual characteristic model applied in the scene before reclassification is equal to or higher than a predetermined threshold value Kth in the scene after reclassification, and the learning efficiency K When it is determined that is equal to or greater than a predetermined threshold value Kth (K ≧ Kth), the process proceeds to step S405 (see FIG. 6).

In step S405, the moving unit ECU (not shown) replaces the reclassified scene with the newly classified scene, updates the scene determination criterion database 10, and ends the process (see FIG. 6). ..

In step S404, when the learning progress determination unit 16 determines in the scene after reclassification that the learning efficiency K by the individual characteristic model applied in the scene before reclassification is less than the predetermined threshold value Kth (K). <Kth), the process proceeds to step S406, and the automatic operation control that has been initially set (default) is performed (see FIG. 6).

[Generation of individual characteristic model]
Next, in the individual characteristic learning unit 14, a process of learning the individual characteristics of the driver (generation of an individual characteristic model) will be described.
FIG. 7 is a flowchart of the individual characteristic learning process.

In step S501, the individual characteristic learning unit 14 determines whether or not the scene in which the moving body (vehicle) is placed is switched, and if it is determined that the scene is switched, the process proceeds to step S502.

As an example of a specific method for determining whether or not there is a scene change in the individual characteristic learning unit 14, when scenes of the same type do not appear consecutively (for example, after a gentle curve scene, an intersection scene). (When) is determined that there is a scene switch (step S501: Yes), and when the same type of scenes appear consecutively (for example, a similar straight line scene appears after the straight line scene). In the case), it is determined that there is no scene switching (step S501: No).

In step S502, the individual characteristic learning unit 14 acquires the environmental information 50 collected in the scene from the past information database 13 and the driver's manual operation information 51, and proceeds to step S503.

In step S503, the individual characteristic learning unit 14 executes learning of individual characteristics (generation of an individual characteristic model). As a result, the individual characteristic learning unit 14 generates an individual characteristic model for estimating the manual operation of the driver (individual) with respect to the scene and the environmental information 50.

If the individual characteristic learning unit 14 determines in step S501 that there is no scene switching (step S501: No), the individual characteristic learning unit 14 proceeds to step S504.

In step S504, the individual characteristic learning unit 14 starts from the time when the individual characteristic learning process (steps S501 to S504) is executed, and whether the mileage L of the moving body (vehicle) is increased from the predetermined threshold value Lth. It is determined whether (L ≧ Lth) or the traveling time H is greater than the predetermined threshold value Hth.

When the individual characteristic learning unit 14 determines that the mileage L of the moving body is greater than the predetermined threshold value Lth (L ≧ Lth), or the travel time H is greater than the predetermined threshold value Hth ( When it is determined that H ≧ Hth) (step S504: Yes), the process proceeds to step S502, the environment information 50 in the scene and the driver's manual operation information 51 are acquired, and it is determined that neither of them has increased. (Step S504: No), the process ends.
As a result, even if the individual characteristic learning unit 14 determines that there is no scene switching, if it is determined that the moving body has passed a predetermined mileage (or a predetermined mileage), the scene switching determination is made. In the same manner as above, an individual characteristic model is generated.

As described above, the individual characteristic learning unit 14 learns the individual characteristics of the driver in a relatively long cycle, such as when the scene is switched, a predetermined mileage is ran, or a predetermined mileage is ran. By doing so, it is possible to prevent the individual characteristic model from being unnecessarily generated (updated) in a short cycle.

[Judgment of completion of individual characteristic learning]
Next, an example of determining whether or not the learning of individual characteristics has been completed will be described in the learning progress determination unit 16.
FIG. 8 is a flowchart of the learning completion determination process of the individual characteristic in the learning progress determination unit 16.

In step S601, the learning progress determination unit 16 determines the current scene in which the moving body (vehicle) is placed. The determination of the scene by the learning progress determination unit 16 is performed by the same process as in steps S201 to S205 in FIG. 3 described above, and thus the description thereof will be omitted here.

In step S602, the learning progress determination unit 16 acquires the individual authentication information of the driver (individual) and the current scene from the past information database 13, and calls the learning result of the individual characteristics in the current scene.

In step S603, the learning progress determination unit 16 determines whether or not the learning result of the individual characteristic (individual characteristic model) exists, and when it is determined that the learning result of the individual characteristic exists (step S603: Yes), The process proceeds to step S604. When the learning progress determination unit 16 determines that the learning result of the individual characteristic does not exist (step S603: No), the process proceeds to step S611, sets the learning progress to "learning incomplete", and ends the process.

In step S604, the learning progress determination unit 16 determines whether or not the moving body is being manually operated by the driver, and if it is determined that the moving body is being manually operated (step S604: Yes), the step S605 is performed. move on. When the learning progress determination unit 16 determines that the moving body is not being manually operated by the driver (step S604: No), the process proceeds to step S611, sets the learning progress to "learning incomplete", and ends the process. do.

In step S605, the learning progress determination unit 16 applies the environmental information 50 acquired in advance to the individual characteristic model generated by the individual characteristic learning unit 14, carries out a running simulation of individual adaptive control, and steps. Proceed to S606.

In step S606, the learning progress determination unit 16 compares the running simulation result (estimated manual operation information) performed in step S605 with the actual manual operation amount (manual operation information 51) of the driver (individual). Then, the difference R is calculated, and the calculated difference R is stored in the past information database 13.

The above-mentioned difference R is a running simulation result such as a steering angle (steering angle amount), an accelerator opening, an acceleration of a moving cheer, a moving body speed, a yaw rate (yaw angle), a positional relationship with an obstacle, and a relative speed. It is the difference from the actual measured value.

Further, the storage of the difference R in the past information database 13 may simply store the accumulation of the difference R, or may store it as a ratio to the travel distance.

In step S607, the learning progress determination unit 16 compares the difference R calculated in step S606 with the predetermined threshold value Rth set for each scene (parameter), and determines whether the difference R is equal to or less than the predetermined threshold value Rth. Judge whether or not.

In step S607, when the learning progress determination unit 16 determines that the difference R is equal to or less than the threshold value Rth (step S607: Yes), the process proceeds to step S608 and determines that the difference R is equal to or less than the threshold value Rth (step S607). S607: No), the process proceeds to step S611, the learning progress is set to "learning incomplete", and the process ends.

In step S607, as an example of the method of setting the predetermined threshold value Rth, a numerical value considered to be an allowable range may be calculated and set in advance. Even if the driver's heart rate increase is obtained from the presence / absence of the (driver) override and the individual's biometric information, and the frequency of the override or heart rate increase occurs is equal to or higher than the predetermined threshold value Rth. good.

In step S608, the learning progress determination unit 16 determines whether or not the past learning progress was in the "learning incomplete" state, and if it is determined that the learning progress is in the "learning incomplete" state (step S608: Yes), the step. If the process proceeds to S609 and it is determined that the state is not in the "learning incomplete" state (step S608: No), the process proceeds to step S610, the learning progress is set to "learning completed", and the process ends.

In step S609, the learning progress determination unit 16 determines whether or not the transition to the learning completion state is permitted. When the learning progress determination unit 16 determines that the transition to the learning completion state is permitted (step S609: Yes), the process proceeds to step S610, sets the learning progress to "learning completed", and performs processing. finish.
On the other hand, when the learning progress determination unit 16 determines that the transition to the learning completed state is not permitted (step S609: No), the process proceeds to step S611, sets the learning progress to "learning incomplete", and processes. To finish.

As an example of the case where the learning progress determination unit 16 determines that the transition to the learning completion state is permitted, it is at the time of switching scenes, and more preferably, at least in the same scene, the scenes are discontinuous. It is preferable that the difference between the simulation result (estimated manual operation information) at two or more data collection opportunities and the actual manual operation information 51 of the driver (individual) is equal to or less than a predetermined threshold value. ..

Here, an example of determining permission to shift to the learning completion state will be described based on the learning progress determined by the learning progress determination unit 16.
FIG. 9 is a diagram illustrating an example of determining permission to shift to the learning completion state based on the learning progress.

In FIG. 9, the scene classification result is shown in the upper part. This scene classification result is a scene determination result (classification result) by the scene determination unit 11.

In the embodiment, two scenes, a curved scene and a straight scene, are classified as the scene classification result, and the curved scene (curved road) and the straight scene (straight road) appear alternately.

The difference R is shown in the lower part of the scene classification result. This difference R is an absolute value of the difference R between the simulation result (estimated manual operation information) of the individual characteristics of the driver estimated by the manual operation estimation unit 15 and the actual manual operation information 51 of the driver.
Instead of the difference R, a value obtained by accumulating the difference R (cumulative difference), or a value obtained by dividing the difference R or the cumulative difference by the mileage may be used, or an average value thereof may be used.

In the embodiment, the accumulation of the difference R between the simulation result and the actual manual operation information 51 of the individual is less than the predetermined threshold value Rth (broken line in the figure) at times t1 to t3, and the threshold value at time t4. It becomes Rth or more.

In the lower part of the difference R, the determination result of the learning progress for each scene (in the embodiment, the curve scene and the straight line scene) is shown.

In the embodiment, when the same scene (for example, a curve scene) occurs twice and the value of the difference R does not exceed a predetermined threshold value Rth in any of the scenes, the determination result of the learning progress is determined to be the completion of learning. ing.
Specifically, the learning progress of the curve scene increases by one step at the time t1 when the first curve scene ends, but does not exceed the threshold value (broken line in the figure) of the learning completion transition permission. Then, the learning progress of the curve scene is further increased by one step at the time t3 when the second curve scene ends, and exceeds the threshold value (broken line in the figure) of the learning completion transition permission.

As a result, as shown in the lower part of the learning progress of FIG. 9, the learning progress determination unit 16 determines that the learning of the individual characteristics in the curve scene is completed at the time t3, and the next curve scene is at the time t4. If it appears after that, individual adaptation control is performed in the curve scene.

On the other hand, the learning progress of the straight line scene increases by one step at the time t2 when the first straight line scene ends, but does not exceed the threshold value (broken line in the figure) of the learning completion transition permission. Then, at the time t4 when the second straight line scene ends, the above-mentioned difference R exceeds the predetermined threshold value Rth, so that the learning progress of the straight line scene does not increase after the time t2, and the learning completion transition permission threshold value. (Dashed line in the figure) does not exceed.

As a result, as shown in the lower part of the learning progress of FIG. 9, the learning progress determination unit 16 does not perform the individual adaptation control in the straight scene even if the next straight scene appears after the time t4. Implement the preset automatic operation control.

After traveling at times t3 to t4 when the difference R exceeds a predetermined threshold value, the learning progress of the straight line scene may be lowered without being left unchanged. Further, the determination of the learning progress does not necessarily have to be stepwise (two steps in the embodiment), and may be continuously changed depending on the magnitude of the difference between the difference R and the predetermined threshold value Rth. .. As a result, it is possible to prevent individual adaptation control from being carried out in an inappropriate learning state such as so-called overfitting, in which the actual individual characteristics are adapted only to a specific scene (place).

[Safety judgment unit]
In the embodiment, in the automatic control device 1, whether or not the individual adaptive control based on the learning result (generated individual characteristic model) in the individual characteristic learning unit 14 is safe in the automatic driving of the moving body (vehicle). Further has a safety determination unit 20 for determining the above.
FIG. 10 is a block diagram illustrating the safety determination unit 20.

As shown in FIG. 10, the safety determination unit 20 includes a safety verification scenario database (DB) 21, a driving simulation unit 22, and a danger event determination unit 23.

The safety verification scenario database 21 contains events that may lead to an accident in various driving conditions and driving environments, such as events that destabilize the behavior of the vehicle and abnormal behavior of other vehicles. Various scenario data of dangerous events that may occur during driving, such as events that may cause a danger such as a collision, are stored.

The above-mentioned events that destabilize the behavior of the vehicle include, for example, when the outputs of various sensors for acquiring the environmental information 50 are unstable, when crosswinds and unevenness of the road surface are large and the steering wheel is easily taken. Can be mentioned.

The scenario data stored in the safety verification scenario database 21 is created in advance based on a predetermined dangerous event. Further, this scenario data is based on the environmental information 50 collected from another moving body (other vehicle) or the like, and an accident actually occurs, a large acceleration is applied to the moving body (other vehicle), or the driver. It is updated by referring to the scene when the heart rate suddenly increases as the biological information of the (individual).

In this way, in the safety verification scenario database 21, the scenario data is updated based on the situation when the dangerous event actually occurs, so that the scenario data in more dangerous events is stored, and various dangerous events are stored. It is possible to correspond to.

The driving simulation unit 22 extracts a scene scenario to be verified by the individual characteristic model (driver) from the safety verification scenario database 21. Then, the driving simulation unit 22 includes the driver's individual characteristic model generated by the individual characteristic learning unit 14, the scene scenario extracted from the safety verification scenario database 21, and the scene determined by the scene determination unit 11 (determination result). Based on the above, a running simulation (simulated running) is carried out.
As a result, the traveling simulation unit 22 carries out a traveling simulation of what kind of driving the individual characteristic model of the driver will perform when a dangerous event occurs in a predetermined scene.

The dangerous event determination unit 23 acquires the result of the driving simulation from the driving simulation unit 22, determines whether or not a dangerous event has occurred or the degree of danger, and uses the determination result as a safety evaluation result in the automatic control device 1. Is output to the control unit 17 (see FIG. 1).

The control unit 17 determines whether or not to implement the individual adaptation control based on the safety evaluation result of the danger event determination unit 23. In this way, the control unit 17 determines whether or not to perform individual adaptive control based on the safety evaluation result, so that, for example, an individual characteristic model created based on learning of the individual characteristics of the driver can be obtained. Even if you have learned the individual characteristics of a driver who cannot respond to dangerous events or who tends to drive dangerously, you can prevent the actual dangerous driving by not executing individual adaptive control. Can be done.

Next, an example of determining the occurrence of a dangerous event by the dangerous event determination unit 23 will be described.
FIG. 11 is a diagram for explaining an example of determining the occurrence of a dangerous event by the dangerous event determination unit, the upper part of the figure is a diagram showing the relationship between the time and the speed of the preceding vehicle, and the lower part of the figure is a diagram showing the relationship between the time and the speed of the preceding vehicle. It is a figure which shows the inter-vehicle distance between a time and a preceding vehicle.
In the lower part of FIG. 11, the case where the dangerous event is appropriately avoided is shown in practice, and the case where the dangerous event cannot be avoided is shown by the broken line.

As shown in the upper part of FIG. 11, for example, the preceding vehicle is traveling on a straight road or the like at a constant vehicle speed until a predetermined time t1, and after the time t1, suddenly starts decelerating. Shown.
This is the scenario data stored in the safety verification scenario database 21, and the dangerous event in this case is so close that the own vehicle may collide with the preceding vehicle (closer than the allowable shortest inter-vehicle distance Lth). )It is to be.

The driving simulation unit 22 is based on the individual characteristic model generated by the individual characteristic learning unit 14, the scene determined by the scene determination unit 11 (determination result), and the scenario data extracted from the safety verification scenario database 21. Perform a running simulation (simulated running).

As a result, as shown in the lower part of FIG. 11, when the own vehicle cannot cope with the sudden deceleration of the preceding vehicle and approaches closer than the predetermined inter-vehicle distance (allowable shortest inter-vehicle distance Lth) (broken line in the lower part of FIG. 11). ), The dangerous event determination unit 23 determines that a dangerous event has occurred, and determines that the individual characteristic model is a model with insufficient safety.

On the other hand, if it is possible to drive the vehicle so that the inter-vehicle distance between the own vehicle and the preceding vehicle does not come closer than the predetermined allowable shortest inter-vehicle distance Lth by vehicle speed control or steering control of the own vehicle (solid line in the lower part of FIG. 11). , The dangerous event determination unit 23 determines that a dangerous event does not occur, and determines that the individual characteristic model is a safe model.

The safety judgment unit 20 performs the above-mentioned safety verification on all scenario data (dangerous events) stored in the safety verification scenario database 21, and verifies the safety of the individual characteristic model in each scene. conduct.

In the above-described embodiment, the case where the automatic control device 1 is applied to the individual adaptive control of the vehicle has been described as an example, but the present invention is not limited to the individual adaptive control of the vehicle, and is not limited to the individual adaptive control of the vehicle. It can also be applied to adaptive control.
For example, it can be applied to individual adaptive control that learns the manual operation of the operator in order to efficiently and automatically drive the drone flight according to the scene such as terrain, weather conditions, and flight purpose.

It can also be used for individual adaptive control to learn manual operations by skilled workers when automating equipment used at manufacturing sites such as factories, construction sites, mining sites, and the like.

As described above, in the embodiment,
(1) An individual authentication unit 12 that authenticates a driver (individual) who operates a moving body (equipment) such as a vehicle, a scene determination unit 11 that determines a scene in which the moving body is placed, and an individual driver. In the individual characteristic learning unit 14 that learns the characteristics, the manual operation estimation unit 15 that estimates the manual operation of the moving body by the driver based on the individual characteristics of the driver, and the predetermined scene determined by the scene determination unit 11. , The estimated manual operation information (estimated manual operation) estimated by the manual operation estimation unit 15 based on the individual characteristics of the driver is compared with the actual manual operation information 51 (manual operation) of the moving body by the driver. , Adapted to the individual characteristics of the driver based on the learning progress determination unit 16 for determining the learning progress in a predetermined scene and the learning progress determination result for each predetermined scene determined by the scene determination unit 11. The configuration includes a control unit 17 that determines whether or not automatic control is performed for each scene.

With this configuration, the automatic control device 1 has a learning progress determination unit 16 that determines the learning progress of the individual characteristics of the driver for each scene determined by the scene determination unit 11. Therefore, the control unit 1 has a learning progress determination unit 16. In 17, it is possible to decide whether or not to implement the individual adaptation control based on the learning progress determined for each scene.

Further, since the learning progress determination unit 16 determines the learning progress for each scene determined by the scene determination unit 11, the control unit 17 determines whether or not to perform individual adaptation control for each scene. It is possible to quickly implement individual adaptation control from a scene with a high degree of learning progress. Therefore, it is possible to prevent the driver from feeling uncomfortable.

(2) The control unit 17 further has a safety determination unit 20 that determines whether or not the moving body can maintain a safe state when the moving body is operated by automatic control adapted to the individual characteristics of the driver. The safety determination unit 20 is configured to perform automatic control adapted to the individual characteristics of the driver only when it is determined that the moving body can maintain a safe state.

With this configuration, the control unit 17 performs individual adaptation control only when it is determined by the safety determination unit 20 that the moving body can maintain a safe state as a result of performing individual adaptation control. Therefore, it is safe. Even if the learning progress is high, the determination unit 20 does not perform individual adaptation control when it is determined that the safety of the moving body cannot be maintained.
Therefore, even if the individual characteristics of the driver who performs dangerous operations on a daily basis are learned, it is possible to prevent the moving body from becoming in a dangerous state by not performing the individual adaptation control.

(3) The scene determination unit 11 determines and controls at least two or more discontinuous first scenes (for example, gentle curve scenes) and second scenes (for example, intersection scenes) included in a predetermined scene. When the learning progress determination unit 16 determines that at least two or more discontinuous first scenes and the learning progress determination results in the second scene are both in the learning completion state, the unit 17 operates. The configuration is such that automatic control is performed according to the individual characteristics of the person.

With this configuration, the learning progress determination unit 16 determines the learning progress in two or more different scenes included in the same predetermined scene, and therefore learns in a plurality of scenes that are almost unchanged in the same predetermined scene. It is possible to prevent the progress from being unnecessarily determined, and it is possible to more reliably determine the learning progress of the predetermined scene.

(4) The scene determination unit 11 determines a third scene (for example, a first straight line scene) and a fourth scene (for example, a second straight line scene) included in a predetermined scene and separated from each other by a predetermined distance. , The control unit 17 is determined by the learning progress determination unit 16 that in a predetermined scene, the determination results of the learning progress in the third scene and the fourth scene, which are separated from each other by a predetermined distance, are both in the learning completed state. In this case, the configuration is such that automatic control is performed according to the individual characteristics of the driver.

With this configuration, the learning progress determination unit 16 determines the learning progress in two or more scenes included in the same predetermined scene and separated from each other by a predetermined distance, so that the positions of the same predetermined scene are almost unchanged. It is possible to prevent the learning progress from being unnecessarily determined in the plurality of scenes of the above, and it is possible to more reliably determine the learning progress of the predetermined scene.

(5) Further has a scene judgment standard database 10 for storing a plurality of scenes in which a moving object is placed, and the scene judgment standard database 10 has a learning progress of a predetermined scene determined by a learning progress judgment unit 16. When the determination result (difference R) of is lower than the predetermined threshold Rth, or when the learning efficiency K of the predetermined scene calculated by the learning progress determination unit 16 is lower than the predetermined threshold Kth, the predetermined The scene (for example, a curve scene) is changed to another scene (for example, a gentle curve scene and a sharp curve scene) different from the predetermined scene (curve scene).

A scene with extremely low learning progress for any driver is likely to have a problem in classifying the scene, and even if such a scene with extremely low learning progress is memorized forever, appropriate learning is performed in the scene. Can't do. In addition, continuing to store such scenes wastes the memory capacity of the database and increases maintenance and management costs.
With this configuration, by reviewing the scenes with low learning progress and reclassifying them into new scenes, it is possible to efficiently perform learning in the reclassified new scenes. In addition, by reclassifying and updating scenes with extremely low learning progress, it is possible to eliminate waste of database memory capacity and optimize maintenance and management costs.

(6) The learning progress determination unit 16 determines the learning progress of other scenes (for example, a gentle curve scene and a steep curve scene) after the change, and determines the learning progress of the predetermined scene (for example, the scene before the change). , Curve scene) The estimated manual operation information (individual characteristic model) used to judge the learning progress is compared with the actual manual operation information 51 of the moving body.
When the learning progress determination result in another scene (for example, a gentle curve scene and a steep curve scene) is equal to or higher than a predetermined threshold, the scene determination criterion database 10 determines the other scene (for example, a gentle curve scene). And a sharp curve scene) is replaced with a predetermined scene (for example, a curve scene) and stored.

With this configuration, it is possible to effectively utilize the estimated manual operation information (individual characteristic model) of the scene before the change, and it is possible to quickly update the scene determination reference database 10.

[Scene match verification unit]
In the above-described embodiment, a case where a predetermined scene having a remarkably low learning efficiency K is subdivided and reclassified into a new scene has been described as an example, but the present invention is not limited to this. For example, two or more subdivided scenes may be grouped together and reclassified into major scenes.
Further, regardless of the learning efficiency K of the scene, if it is determined that one scene and another scene match, the scene may be reclassified into one scene.

Next, the scene matching degree verification unit 30 of the automatic control device 1 according to another embodiment will be described.
FIG. 12 is a block diagram illustrating a scene matching degree verification unit 30.

As shown in FIG. 12, the scene matching degree verification unit 30 has a driving scene scenario database 31 and a reclassification determination unit 32.

The driving scene scenario database 31 stores a plurality of driving scene scenarios including each scene stored in the scene determination standard database 10.
Here, the traveling scene scenario is information (scenario) including a predetermined scene of a road on which a moving body such as a vehicle travels and environmental information when the moving body travels in the predetermined scene.

The driving simulation unit 22 reads all the driving scene scenarios from the driving scene scenario database 31, and also reads the individual characteristic model corresponding to each scene included in the driving scene scenario generated by the individual characteristic learning unit 14.

In the running simulation unit 22, for example, a running simulation (simulated running) is executed by fitting all the read individual characteristic models one by one to one of all the read running scene scenarios. Then, the running simulation unit 22 executes the same processing for all the read running scene scenarios.

In the reclassification determination unit 32, as a result of the running simulation (simulated running) in the running simulation unit 22, a plurality of running scene scenarios (for example, a straight scene scenario with a gentle slope and a steep slope) in which the running simulations match (or are similar) Straight-line scene scenarios) are combined into one to create a new driving scene scenario (for example, straight-line scene scenario).

As a result, the number of driving scene scenarios (scenes stored in the scene determination reference database 10) stored in the driving scene scenario database 31 can be reduced. Therefore, the memory area of the scene determination reference database 10 for storing the scene can be reduced, and it is possible to prevent an increase in manufacturing and management costs of the automatic control device 1 due to the enormous memory area.

<Second Embodiment>
Next, the automatic control device 1A according to the second embodiment will be described.
When the automatic control device 1A according to the second embodiment is applied to learning the individual characteristics of the long-term care system in the long-term care system of a person in need of long-term care (hereinafter referred to as a long-term care recipient or a long-term care target person). Will be illustrated and described.
FIG. 13 is a block diagram illustrating the automatic control device 1A according to the second embodiment. The same configuration as that of the above-described embodiment will be assigned the same number and will be described as necessary.

[Automatic control device]
As shown in FIG. 13, in the automatic control device 1A, the care-requiring person environment information 53 is input to the scene determination unit 11.
The care-requiring person environmental information 53 includes external environmental information (weather, position, object information that the care-requiring person is interested in, etc.) around the care-requiring person (watching target person), and biological information of the care-requiring person. Includes internal environment information such as and movement speed.

The care-requiring person environmental information 53 may be collected by a camera installed in the facility where the care-requiring person enters or on the infrastructure side, or may be collected by a wearable terminal worn by the care-requiring person.

Caregiver determination information 54 is input to the past information database 13. The caregiver judgment information 54 is information on the degree of risk determined by the caregiver (the person who cares for the care recipient) by checking the surrounding situation, and is stored in a predetermined memory address of the past information database 13. It is assumed that the caregiver judgment information 54 determines that the risk level is 0 (zero) when there is no input by the caregiver (judgment result of the risk level).

Further, the automatic control device 1A has a dangerous event determination unit 18.
The danger event determination unit 18 is input with the current environment information 53 for the person requiring long-term care and the individual characteristic model generated by the individual characteristic learning unit 14. The dangerous event determination unit 18 determines the presence or absence of a dangerous event and estimates the probability of occurrence of the dangerous event from the input care-requiring person environmental information 53 and the individual characteristic model.

Here, the dangerous event means, for example, a situation in which a person requiring long-term care approaches a road with heavy traffic or is interested in an object at risk of injury or falling.

The learning progress determination unit 16 compares the determination result of the danger event in the danger event determination unit 18 with the caregiver determination information 54 (determination result of the caregiver's risk degree), and the smaller the difference R, the more the said. It is judged that the learning progress of individual characteristics in the scene is high.

Then, the control unit 17 executes individual adaptation science control from the scene determined to have a high learning progress, and the scene determined to have a low learning progress is determined to have a high learning progress. No individual adaptation control is performed.

[Individual adaptation control]
Next, the flow of individual adaptive control in the automatic control device 1A according to the second embodiment will be described.
FIG. 14 is a flowchart illustrating the flow of individual adaptation control.

In step S701, the overall control unit (not shown) of the watching care system determines whether or not control of the watching care system is permitted as a trigger for collecting data and learning individual characteristics.

When the overall control unit (not shown) determines that control is permitted (step S701: Yes), proceeds to step S702, and determines that control is not permitted (step S701: No). End the process.
The determination of the control permission may be made based on the ON / OFF of the activation switch of the caregiver's watching care system, or may be made based on the ON / OFF of the power supply of the watching care system.

In step S702, the overall control unit (not shown) of the watching care system acquires the environmental information around the care recipient (care recipient environmental information 53) and transmits it to the past information database 13.

In step S703, the control unit 17 determines whether or not the individual adaptation determination is permitted. When the control unit 17 determines that the individual adaptation determination is permitted (step S703: Yes), the process proceeds to step S704, and when it is determined that the individual adaptation determination is not permitted (step S703: No), the step Proceed to S706, and perform the preset initial setting (default) control.

In step S704, the learning progress determination unit 16 determines whether or not the learning of the individual characteristics is completed in the current scene. When the learning progress determination unit 16 determines that the learning progress is higher than the predetermined threshold value in the current scene (step S704: Yes), the learning progress determination unit 16 proceeds to step S705 and determines that the learning progress is lower than the predetermined threshold value. If it is determined (step S704: No), the process proceeds to step S706, and the control that has been initialized (default) in advance is executed.

Then, in step S705, the control unit 17 executes individual adaptation control and ends the process.

[Past information database update process]
Next, the update process of the past information database 13 will be described.
FIG. 15 is a flowchart of the update process of the past information database 13.

In step S801, the individual authentication unit 12 acquires the individual authentication information 52 of the current care recipient.

In step S802, the individual authentication unit 12 determines whether or not the current individual authentication information 52 of the person requiring long-term care acquired in step S801 is already stored in the past information database 13. When the individual authentication unit 12 determines that the individual authentication information 52 is already stored in the past information database 13 (step S802: Yes), the individual authentication unit 12 proceeds to step S803.

In step S803, the individual authentication unit 12 acquires the memory address on the past information database 13 in which the individual authentication information 52 of the person requiring long-term care is stored, and proceeds to step S804.

On the other hand, when the individual authentication unit 12 determines that the individual authentication information 52 of the person requiring long-term care is not stored in the past information database 13 (step S802: No), the individual authentication unit 12 proceeds to step S808.

In step S808, the individual authentication unit 12 sets a new memory address for storing the current individual authentication information 52 of the person requiring long-term care on the past information database 13, and proceeds to step S804.

In step S804, the scene determination unit 11 acquires the care-requiring person environment information 53 and proceeds to step S805.

In step S805, the scene determination unit 11 determines the scene based on the parameters necessary for determining the scene acquired from the scene determination reference database 10 and the care-requiring person environment information 53 acquired in step S804.

In step S806, the overall control unit (not shown) of the automatic control device 1A acquires the caregiver risk determination information (caregiver determination information 54).

In step S807, the overall control unit (not shown) of the automatic control device 1A transmits the acquired care-requiring environment information 53 and the caregiver judgment information 54 to the past information database 13 and the danger event judgment unit 18, respectively. And end the process.

[Weighting]
In the above-mentioned learning of the long-term care-requiring person's dangerous event, when calculating the difference between the results of determining the dangerous event by the learning progress determination unit 16, an alert is issued to the long-term care person by the watching care system to see if the caregiver is nearby. The weight may be weighted based on the difference from the caregiver's judgment at the time of the event.
As a result, when the caregiver cannot recognize the dangerous event, it is possible to prevent the watching care system from mistakenly learning that it was not dangerous.

Next, a process of weighting the difference between the caregiver's risk judgment and the risk judgment by the watching care system will be described.
FIG. 16 is a flowchart of the weight determination process.

First, in step S901, the overall control unit (not shown) of the watching care system determines whether or not the caregiver has input the judgment of the dangerous event. When it is determined that the dangerous event has been input (step S901: Yes), the process proceeds to step S902 and the weight W1 is selected.

The weight W1 selected in step S901 is set in the range of 0 to 1, and is reduced according to the elapsed time from the occurrence of the dangerous event specified by the caregiver to the input of the judgment of the dangerous event, for example. May be good. In addition, it may be set individually for each caregiver who determines the degree of risk.

In step S901, when the overall control unit (not shown) of the watching care system determines that there is no input of a dangerous event (step S901: No), the caregiver proceeds to step S904 and the caregiver is a care recipient (watching target). Determine whether or not the person is conscious of the person).

In step S904, when the overall control unit (not shown) of the watching care system determines that the caregiver is conscious of the person requiring long-term care (watching target person) (step S904: Yes), the process proceeds to step S905. The weight W2 is selected.

Here, the weight W2 selected in step S905 is set to a value smaller than the weight W1 selected in step S902. The weight W2 can be calculated, for example, by multiplying the weight W1 by a coefficient in the range of 0 to 1.

Further, in step S904, various sensors for acquiring the care-requiring person environmental information 53 are used to determine whether or not the caregiver is conscious, and for example, the care-requiring person's eyes, face, and body are the care-requiring person. If it is directed in the direction of (the person to be watched over), it can be determined that the consciousness is directed.

In step S904, when the overall control unit (not shown) of the watching care system determines that the caregiver is not paying attention to the care recipient (watching target person) (step S904: No), the process proceeds to step S906. Then, it is determined whether or not the caregiver is in a place where the care recipient (person to be watched over) can be seen.

In step S906, when the overall control unit (not shown) of the watching care system determines that the care recipient (the person to be watched) can be visually recognized (step S906: Yes), the care proceeds to step S907 and the weight is increased. W3 is selected.

Here, the weight W3 selected in step S907 is set to a value smaller than the weight W2 selected in step S905. For example, the weight W3 can be calculated by multiplying the weight W2 by a coefficient in the range of 0 to 1, which decreases according to the distance between the caregiver and the care recipient (the person to be watched over).

In step S906, when the overall control unit (not shown) of the watching care system determines that the caregiver is in a position where the care recipient (watching target person) cannot be seen (step S906: No), the process proceeds to step S908. The weight W4 is selected.

Here, the weight W4 selected in step S908 is set to a value smaller than the weight W1 selected in step S902. The weight W4 can be calculated by multiplying the weight W1 by a coefficient in the range of 0 to 1.

After the weights W1 to W4 are set, in step S903, in step S902, step S905, step S907, and step S908, the difference between the risk judgment result by the caregiver and the risk judgment result by the watching care system is calculated. Multiply the selected weights W1, W2, W3, W4.

In step S909, the cumulative value of the weighted difference (for example, the average value of each scene) is calculated, and the process ends.

The judgment result of the dangerous event in the above-mentioned watching care system is not only used by human beings to appropriately confirm the state of the person requiring long-term care (person to be watched over), but also accompanies the person requiring long-term care and a dangerous event occurs. Then, when it is predicted, the estimation result of this dangerous event may be output to the robot that works to avoid it.

As described above, in the embodiment,
In the past, when the care recipient was not accompanied by the care recipient (person to be watched over), it was necessary to restrict the behavior of the care recipient in order to prevent the occurrence of dangerous events. However, the above-mentioned watching care system can determine whether or not the behavior is within a safe range suitable for the individual ability of the care recipient on behalf of the caregiver.
As a result, it is possible to improve the quality of life of the long-term care recipient by increasing the range of action and possibility of the long-term care recipient while reducing the burden on the long-term care recipient.

The automatic control device 1 according to the first embodiment described above may be combined with the automatic control device 1A according to the second embodiment, or may be combined with other embodiments.

1, 1A: Automatic control device, 10: Scene judgment standard database, 11: Scene judgment unit, 12: Individual authentication unit, 13: Past information database, 14: Individual characteristic learning unit, 15: Manual operation estimation unit, 16: Learning Progress judgment unit, 17: Control unit, 20: Safety judgment unit, 21: Safety verification scenario database, 22: Driving simulation unit, 23: Dangerous event judgment unit, 30: Scene matching degree verification unit, 31: Driving scene scenario Database, 32: Reclassification judgment unit, 50: Environmental information, 51: Manual operation information, 52: Individual authentication information

Claims (8)

An individual authentication unit that authenticates an individual who operates a device,
A scene determination unit that determines the scene in which the device is placed, and
The individual characteristic learning unit that learns the individual characteristics of the individual,
A manual operation estimation unit that estimates the manual operation of the device by the individual based on the individual characteristics of the individual, and a manual operation estimation unit.
In a predetermined scene determined by the scene determination unit, the estimated manual operation estimated by the manual operation estimation unit based on the individual characteristics of the individual is compared with the actual manual operation of the device by the individual. A learning progress determination unit that determines the learning progress in the predetermined scene,
Based on the determination result of the learning progress for each of the predetermined scenes determined by the scene determination unit, the control unit determines the execution or non-execution of the automatic control adapted to the individual characteristics of the individual for each scene. When,
Have a, a scene determination criteria database for storing a plurality of scenes in which the device is located,
The scene judgment criteria database is
When the learning progress determination result of the predetermined scene determined by the learning progress determination unit is lower than the predetermined threshold value, or the learning efficiency of the predetermined scene calculated by the learning progress determination unit is high. When the threshold value is lower than the predetermined threshold value, the predetermined scene is changed to another scene different from the predetermined scene.
The learning progress determination unit uses the determination of the learning progress of the other scene after the change to determine the learning progress of the predetermined scene before the change, and the device. Performed by comparing with the actual manual operation of
The scene determination reference database is an automatic control device that replaces the other scene with one scene and stores the other scene when the determination result of the learning progress in the other scene is equal to or more than a predetermined threshold value. An individual authentication unit that authenticates an individual who operates a device,
A scene determination unit that determines the scene in which the device is placed, and
The individual characteristic learning unit that learns the individual characteristics of the individual,
A manual operation estimation unit that estimates the manual operation of the device by the individual based on the individual characteristics of the individual, and a manual operation estimation unit.
In a predetermined scene determined by the scene determination unit, the estimated manual operation estimated by the manual operation estimation unit based on the individual characteristics of the individual is compared with the actual manual operation of the device by the individual. A learning progress determination unit that determines the learning progress in the predetermined scene,
Based on the determination result of the learning progress for each of the predetermined scenes determined by the scene determination unit, the control unit determines the execution or non-execution of the automatic control adapted to the individual characteristics of the individual for each scene. When,
A scene judgment standard database that stores multiple scenes in which the device is placed, and
It has a scene matching degree verification unit for verifying the matching degree between one scene and another scene among a plurality of scenes included in the scene determination standard database.
The scene matching degree verification unit
Judgment of the learning progress degree determined by the learning progress determination unit by applying one estimated manual operation estimated by the manual operation estimation unit from the individual characteristics of the individual in the one scene to the other scene. Results and
Judgment of the learning progress degree determined by the learning progress determination unit by applying another estimated manual operation estimated by the manual operation estimation unit from the individual characteristics of the individual in the other scene to the one scene. Compare the result with
An automatic control device that updates the scene determination reference database by collecting the one scene and the other scene into a common scene when the result of the comparison is smaller than a predetermined threshold value. Further, it has a safety determination unit for determining whether or not the device can maintain a safe state when the device is operated by automatic control adapted to the individual characteristics of the individual.
The control unit
The automatic control device according to claim 1 or 2 , wherein automatic control adapted to the individual characteristics of the individual is performed only when the safety determination unit determines that the device can maintain a safe state. The scene determination unit determines at least two or more discontinuous first scenes and second scenes included in the predetermined scene, and determines the second scene.
The control unit
When the learning progress determination unit determines that at least two or more discontinuous first scenes and the learning progress determination results in the second scene are both in the learning completion state, the individual The automatic control device according to claim 3 , wherein automatic control adapted to the individual characteristics of the above is performed. The scene determination unit determines the third scene and the fourth scene included in the predetermined scene and separated from each other by a predetermined distance.
The control unit
When the learning progress determination unit determines that in the predetermined scene, the determination results of the learning progress in the third scene and the fourth scene, which are separated from each other by a predetermined distance, are both in the learning completed state. The automatic control device according to claim 3 , wherein automatic control adapted to the individual characteristics of the individual is performed.
Control device. An individual authentication process that authenticates an individual who operates a device,
A scene determination process for determining the scene in which the device is placed, and
An individual characteristic learning process for learning the individual characteristics of the individual, and
A manual operation estimation step for estimating the manual operation of the device by the individual based on the individual characteristics of the individual, and a manual operation estimation step.
In a predetermined scene determined in the scene determination step, the estimated manual operation estimated in the manual operation estimation step based on the individual characteristics of the individual is compared with the actual manual operation of the device by the individual. A learning progress determination step for determining the learning progress in the predetermined scene, and
Based on the determination result of the learning progress for each of the predetermined scenes determined in the scene determination step, the control step of determining the execution or non-execution of the automatic control adapted to the individual characteristics of the individual for each scene. When,
Among the plurality of scenes included in the scene determination reference database for storing a plurality of scenes in which the equipment is located, possess a scene matching degree verification step of verifying the coincidence of the first scene and another scene, a,
The scene matching degree verification step is
Judgment of the learning progress degree determined by the learning progress determination step by applying one estimated manual operation estimated by the manual operation estimation step from the individual characteristics of the individual in the one scene to the other scenes. Results and
Judgment of the learning progress degree determined by the learning progress determination step by applying another estimated manual operation estimated by the manual operation estimation step from the individual characteristics of the individual in the other scene to the one scene. A comparison process that compares the results,
Result of the comparison is smaller than a predetermined threshold, the automatic control system for chromatic and a updating step of updating the scene determination reference database together said as one of the scene and the other scenes common scenes Control method. The scene determination step determines at least two or more discontinuous first scenes and second scenes included in the predetermined scene, and determines the second scene.
The control step is
When the learning progress determination step determines that at least two or more discontinuous first scenes and the learning progress determination results in the second scene are both in the learning completion state, the individual The control method of the automatic control device according to claim 6 , wherein automatic control adapted to the individual characteristics of the above is performed. In the scene determination step, the third scene and the fourth scene included in the predetermined scene and separated from each other by a predetermined distance are determined.
The control step is
When it is determined by the learning progress determination step that in the predetermined scene, the determination results of the learning progress in the third scene and the fourth scene, which are separated from each other by a predetermined distance, are both in the learning completed state. The control method of the automatic control device according to claim 6 , wherein automatic control adapted to the individual characteristics of the individual is carried out.

Images