WO2019207766A1

WO2019207766A1 - Control device and control method

Info

Publication number: WO2019207766A1
Application number: PCT/JP2018/017207
Authority: WO
Inventors: 中川　慎二
Original assignee: 株式会社日立製作所
Priority date: 2018-04-27
Filing date: 2018-04-27
Publication date: 2019-10-31

Abstract

The present invention performs an abnormality detection on the basis of suitable parameters according to an operation condition or an environment condition of an apparatus, which is an object to be monitored, so as to suitably perform the abnormality detection on the apparatus and reduce a calculation load of a control device according to an increase in parameters for detecting the abnormality. A control device 1, which controls an automatic driving vehicle 5, is configured to have: a normal operation determination unit 4 which determines an operation state of the automatic driving vehicle 5 to be normal, when all parameter values relating to a control, which are transmitted from the automatic driving vehicle 5 on the basis of an operation state of the automatic driving vehicle 5, are included in a normal range that is regulated with a plurality of normal range regulation parameters relating to the control of the automatic driving vehicle 5; a parameter learning unit 2 which learns the normal range regulation parameters that regulate the normal range on the basis of the operation condition or the environmental condition of the automatic driving vehicle 5; and a parameter changing unit which changes or selects the normal range regulation parameters on the basis of the operation condition or the environment condition on operation of the automatic driving vehicle 5.

Description

Control apparatus and control method

The present invention relates to a control device and a control method.

In Patent Literature 1, a feature information value extracted as a characteristic value from acquired information acquired as a value indicating the operating state of a device to be monitored is used to construct a learning parameter obtained in advance. The first learning unit that updates the learning space by reconstructing the learning parameters of the learning space and the feature information value extracted from the acquired information before the update using the feature information value is performed. If the detected value indicating the degree of deviation from the learning space is greater than or equal to the first threshold or exceeds the first threshold, information used for learning from the acquired information from which the feature information value is extracted A selection unit that causes a selection executor or a learning agent to select appropriate acquisition information as a feature information, and feature information extracted as a characteristic value from the acquisition information selected by the selection executor via the selection unit An abnormality determination device comprising: a second learning unit that uses a value to reconfigure a learning parameter in a learning space before being updated by the first learning unit, thereby updating the learning space. Yes.

Japanese Unexamined Patent Publication No. 2016-173682

However, in recent years, with the sophistication and complexity of devices to be monitored, the number of parameters for detecting an abnormality of the device has increased, and the amount of calculation of abnormality detection based on the parameters by the abnormality determination device is enormous. turn into. On the other hand, the abnormality determination device disclosed in Patent Literature 1 does not learn a parameter that is a target of abnormality detection according to an operation condition or an environmental condition of a device that is a monitoring target. Moreover, the parameter which becomes the object of abnormality detection is not changed or selected according to the operating condition or the environmental condition, and the disclosure or suggestion thereof is not made. As a result, in Patent Document 1, abnormality detection based on appropriate parameters according to the operating conditions and environmental conditions of the device to be monitored is not performed, and abnormality detection cannot be performed appropriately. In addition, it is not possible to reduce the calculation load of the abnormality determination device due to an increase in parameters for detecting abnormality.

Therefore, the present invention appropriately performs abnormality detection by performing abnormality detection based on appropriate parameters according to the operating conditions and environmental conditions of the device to be monitored, and abnormality determination accompanying an increase in parameters for detecting abnormality. The purpose is to reduce the calculation load of the apparatus.

In order to solve the above-mentioned problem, a control device for controlling a controlled device, wherein the controlled device is within a normal range defined by a plurality of normal range defining parameters related to the control of the controlled device based on the operating state of the controlled device. When all of the parameter values related to control transmitted from the control device are included, based on the normal operation determination unit that determines that the operation state of the controlled device is normal and the operation condition or environmental condition of the controlled device A parameter learning unit that learns a normal range defining parameter that defines a normal range; and a parameter changing unit that changes or selects a normal range defining parameter based on an operating condition or an environmental condition of the controlled device during operation of the controlled device It was set as the structure which has.

According to the present invention, based on the operation condition or the environmental condition of the controlled device, the normal range defining parameter that defines the predetermined range for determining whether or not the operation is normal is learned in advance, and the operation condition is set when the control device is in operation. Alternatively, since the normal range defining parameter for determining whether or not the normal operation is based on the environmental condition is changed or selected, the target range is determined based on the normal range defining parameter appropriate to the operating condition or the environmental condition of the controlled device. Abnormality detection of the operating state of the control device can be performed appropriately, and the performance of abnormality detection can be improved. Further, since the normal range defining parameter necessary for abnormality detection is used, the amount of arithmetic processing of the control device can be optimized.

It is a block diagram explaining the functional structure of the control apparatus concerning this invention. It is a block diagram explaining the hardware constitutions of a control apparatus. It is a block diagram at the time of applying a control apparatus to control of an autonomous driving vehicle. It is a block diagram explaining the function structure of a parameter learning part. It is a block diagram explaining the function structure of a parameter change part. It is a block diagram explaining the functional structure of a normal operation judgment part. It is a block diagram explaining the functional structure of the normal operation | movement judgment part concerning 2nd Embodiment. It is a block diagram explaining the functional structure of the normal operation judgment part concerning 3rd Embodiment. It is a block diagram explaining the functional structure of the parameter learning part concerning 4th Embodiment. It is a block diagram explaining the functional structure of the parameter learning part concerning 5th Embodiment. It is a block diagram explaining the functional structure of the control apparatus concerning 6th Embodiment. It is a block diagram at the time of applying the control apparatus concerning 6th Embodiment to an autonomous driving vehicle. It is a block diagram explaining the functional structure of the parameter learning part concerning 6th Embodiment. It is a block diagram explaining the functional structure of the parameter change part concerning 6th Embodiment. It is a block diagram explaining the functional structure of the normal operation | movement judgment part concerning 6th Embodiment. It is a block diagram at the time of applying the control apparatus concerning a 7th embodiment to control of a robot. It is a block diagram at the time of applying the control apparatus concerning 8th Embodiment to control of a flying body.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[First Embodiment]
First, the control device 1 according to the first embodiment of the present invention will be described.
FIG. 1 is a block diagram illustrating a functional configuration of the control device 1 according to the embodiment.

<Functional configuration of control device>
As shown in FIG. 1, the control device 1 includes a parameter learning unit 2, a parameter change unit 3, and a normal operation determination unit 4.

The parameter learning unit 2 operates the controlled device based on the operating condition (hereinafter also referred to as driving condition) and the environmental condition of the controlled device controlled by the control device 1 (for example, the automatic driving vehicle 5 shown in FIG. 3). The normal range defining parameter that defines the normal range (normal range) is learned, and the learned normal range defining parameter is transmitted to the parameter changing unit 3. An example of detailed processing in the parameter learning unit 2 will be described with reference to FIG.

The parameter changing unit 3 changes or selects a plurality of normal range defining parameters that define the normal range learned by the parameter learning unit 2, and transmits the normal range defining parameters after the change or selection to the normal operation determining unit 4. An example of detailed processing in the parameter changing unit 3 will be described with reference to FIG.

The normal operation determining unit 4 operates when the controlled device operates when the parameter values related to the control of the controlled device are all included in the normal range defined by the normal range defining parameter changed or selected by the parameter changing unit 3. If the state is determined to be normal and at least some of the parameter values are not included, it is determined that the operating state of the controlled device is abnormal and an abnormal flag is output. An example of detailed processing in the normal operation determination unit 4 will be described with reference to FIG.

<Hardware configuration of control device>
The above-described functions of the parameter learning unit 2, the parameter changing unit 3, and the normal operation determining unit 4 of the control device 1 are exhibited when the CPU 12 described later executes a control program stored in the ROM 13. Next, the hardware configuration of the control device 1 will be described.
FIG. 2 is a block diagram illustrating a hardware configuration of the control device 1.

The control device 1 includes a storage device 11, a CPU (Central Processing Unit) 12, a ROM (Read Only Memory) 13, a RAM (Random Access Memory) 14, an input circuit 16, an input / output port 17, and an output circuit. 18. In each of these devices, information is transmitted / received via the data bus 15.

The input circuit 16 receives signals from various sensors provided in the controlled device (for example, the automatic driving vehicle 5 shown in FIG. 3). For example, when the control target of the control device 1 is the autonomous driving vehicle 5, the driving conditions (for example, vehicle speed, acceleration, etc.) of the autonomous driving vehicle 5 and the surrounding environmental conditions (for example, temperature, humidity, etc.) are various sensors. The signals detected by these various sensors are input to the input circuit 16 and processed. The input signal after being processed by the input circuit 16 is transmitted to the input / output port 17. The input signal transmitted to the input / output port 17 is stored in the RAM 14 or the storage device 11 via the data bus 15.

Note that the processing and functions described above are performed by the CPU 12 executing a control program stored in the ROM 13, but the control program may be stored in the storage device 11. Even in this way, the CPU 12 executes the control program stored in the storage device 11 so that the processing and functions described above are performed. At that time, the CPU 12 appropriately reads out the value stored in the RAM 14 or the storage device 11 and performs the calculation.

As a result of the calculation by the CPU 12, information (value) transmitted to the outside of the control device 1 is transmitted to the input / output port 17 via the data bus 15, and then the information transmitted to the input / output port 17 is output as an output signal. It is transmitted to the circuit 18. The output signal transmitted to the output circuit 18 is transmitted to the outside as an external signal. Here, for example, the signal to the outside refers to an abnormality flag that informs the abnormality of the operation state of the controlled device.

<Example of automatic driving vehicle control by control device>
Next, the case where the control device 1 is applied to the control device of the autonomous driving vehicle 5 will be described as an example.
FIG. 3 is a block diagram when the control device is applied to an autonomous driving vehicle.
FIG. 4 is a block diagram illustrating a functional configuration of the parameter learning unit.
FIG. 5 is a block diagram illustrating a functional configuration of the parameter changing unit.
FIG. 6 is a block diagram illustrating a functional configuration of the normal operation determination unit.

As shown in FIG. 3, the control device 1 is connected to an autonomous driving vehicle 5 that is a controlled device. From the autonomous driving vehicle 5, driving conditions (for example, vehicle speed, acceleration, etc.) and environmental conditions (for example, air temperature, Humidity) and parameters related to control for controlling the autonomous driving vehicle 5 (for example, target throttle opening, target fuel injection amount, target yaw rate, etc.) are transmitted to the control device 1.

<Parameter learning unit>
As shown in FIG. 4, the parameter learning unit 2 of the control device 1 defines a normal range that defines the normal range of the operating state of the automatic driving vehicle 5 from the values of the driving condition and the environmental condition transmitted from the automatic driving vehicle 5. A normal range defining parameter of a dimension in which the difference e between the maximum value and the minimum value of each dimension of the parameter is equal to or greater than a predetermined threshold K1 is extracted and stored in the storage device 11. Here, the dimension means the number of normal range defining parameters. The operating conditions and the environmental conditions may be represented by numbers corresponding to the conditions. Although FIG. 4 illustrates the case where the operating conditions and the environmental conditions are three conditions A, B, and C, the number of operating conditions and environmental conditions is not limited to this.

<Parameter change part>
As shown in FIG. 5, the parameter changing unit 3 learned by the parameter learning unit 2 in accordance with the driving conditions and environmental conditions (represented by numbers corresponding to the respective conditions) transmitted from the autonomous driving vehicle 5. Changing or selecting a predetermined normal range defining parameter from among a plurality of normal range defining parameters defining the normal range. For example, the parameter changing unit 3 selects normal range defining parameters α1 and α2 necessary for a predetermined operation of the controlled device from among a plurality of normal range defining parameters. Note that FIG. 5 illustrates the case where the operating conditions and the environmental conditions are three conditions A, B, and C, but the number of operating conditions and environmental conditions is not limited to this.

<Normal operation determination unit>
As shown in FIG. 6, the normal operation determining unit 4 is within the normal range defined by the normal range defining parameters (in the embodiment, normal range defining parameters α1 and α2) changed or selected by the parameter changing unit 3. If all the parameter values related to the control transmitted from the automatic driving vehicle 5 are present, it is determined that the operation state of the automatic driving vehicle 5 is normal, and if at least some of the parameter values are not within the normal range, the automatic driving vehicle 5 is judged to be abnormal, and an abnormal flag is output. The normal range is defined (learned) in advance by the upper limit value and the lower limit value of the normal range defining parameter relating to normal control.

Specifically, the normal operation determination unit 4 performs the following processes (1) and (2).
(1) If all parameter values relating to control are within the normal range defined by the normal range defining parameters α1 and α2, it is determined that the operation state of the automatic driving vehicle 5 is normal, and the value of the abnormality flag is set to 0. Set to (zero).
(2) In other cases (at least, when there are no parameter values related to control within the normal range), the operation state of the autonomous driving vehicle 5 is determined to be abnormal, and the value of the abnormality flag is set to 1. To do.
6 illustrates the case where the normal range is defined in two dimensions of the normal range defining parameters α1 and α2, the normal range may be defined in N dimensions (for example, three dimensions or more). .

According to the above-described embodiment, the control device 1 learns in advance the normal range defining parameter that defines the normal range based on at least the driving condition and the environmental condition of the autonomous driving vehicle 5 in the control of the autonomous driving vehicle 5. Change or select a plurality of normal range defining parameters that define the normal range based on operating conditions or environmental conditions, and all the parameters related to control are within the normal range defined by the changed or selected normal range defining parameters If it is included, it can be determined that the operation state of the autonomous driving vehicle 5 is normal. The normal range is defined by an upper limit value and a lower limit value of normal range defining parameters relating to normal control. As a result, the control device 1 can appropriately detect the abnormality of the operation state of the automatic driving vehicle 5 based only on the proper normal range defining parameter corresponding to the driving condition and the environmental condition of the automatic driving vehicle 5. The detection performance can be improved. Further, since the control device 1 uses only normal range defining parameters necessary for abnormality detection, it is possible to reduce the amount of arithmetic processing of the control device 1 and to optimize the calculation load.

As described above, in the embodiment,
(1) The control device 1 that controls the autonomous driving vehicle 5 (controlled device), and the operation of the autonomous driving vehicle 5 is within a normal range defined by a plurality of normal range defining parameters related to the control of the autonomous driving vehicle 5. A normal operation determination unit 4 that determines that the operation state of the automatic driving vehicle 5 is normal when all of the parameter values related to control transmitted from the automatic driving vehicle 5 based on the state are included; 5 based on the operation condition or environmental condition of the automatic driving vehicle 5 and the parameter learning unit 2 that learns the normal range defining parameter that defines the normal range based on the operation condition or environmental condition of the automatic driving vehicle 5 And a parameter changing unit for changing or selecting the normal range defining parameter.

If comprised in this way, the control apparatus 1 can perform the abnormality detection of the operating state of the automatic driving vehicle 5 appropriately only based on the appropriate normal range prescription | regulation parameter according to the driving condition and environmental condition of the automatic driving vehicle 5. And the abnormality detection performance can be improved. In addition, since the control device 1 uses only normal range defining parameters that contribute to abnormality detection in the driving conditions and environmental conditions of the autonomous driving vehicle 5, the amount of arithmetic processing of the control device 1 is reduced and the calculation load is optimized. Can be planned.

(2) The normal operation determination unit 4 is configured to define the normal range defined by the normal range defining parameter by the upper limit value and the lower limit value of the normal range defining parameter.

If comprised in this way, the judgment whether the parameter regarding control of the automatic driving vehicle 5 is normal can be judged appropriately based on the normal range prescribed | regulated by the upper limit value and lower limit value of a normal range prescription | regulation parameter. .

(3) Further, the parameter learning unit 2 learns a normal range defining parameter of a dimension in which the difference e between the maximum value and the minimum value in each dimension of the normal range defining parameter value that defines the normal range is greater than or equal to a predetermined threshold K1. The configuration.

If comprised in this way, the parameter learning part 2 of the control apparatus 1 will learn (extract) the normal range prescription | regulation parameter from which the difference e of the upper limit of a normal range prescription | regulation parameter and a lower limit becomes more than threshold value K1. As a result, the parameter learning unit 2 extracts only normal range defining parameters having a predetermined width (difference e) from a plurality of normal range defining parameters that appear in the control of the autonomous driving vehicle 5. Therefore, in the control of the autonomous driving vehicle 5, it is possible to extract only the normal range defining parameter having a high contribution of abnormality determination.
<Second Embodiment>
Next, a control device according to a second embodiment of the present invention will be described.
FIG. 7 is a block diagram illustrating a functional configuration of a normal operation determination unit 4A according to the second embodiment. In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As shown in FIG. 7, the normal operation determination unit 4A determines a normal range defined by the normal range defining parameters as a plurality of normal range defining parameters (α1, α2) related to control of the controlled device (for example, the autonomous driving vehicle 5). ) Define (learn) based on a function (approximation function) that approximates the distribution of values. Then, the normal operation determination unit 4A determines whether or not it is normal based on the following processes (1) and (2).
(1) If all the parameter values relating to the control are on the approximate function defined by the normal range defining parameters α1 and α2 (within the normal range), it is determined that the operation state of the automatic driving vehicle 5 is normal and abnormal. Set the value of the flag to 0 (zero).
(2) In other cases (at least, when there are no values of parameters related to control on the approximate function), the operation state of the autonomous driving vehicle 5 is determined to be abnormal, and the value of the abnormality flag is set to 1. To do.
Note that the normal range may be a range having a range of ± α with respect to the approximate function (see the dotted line in FIG. 6). In FIG. 7, the case where the normal range is defined in two dimensions of the normal range defining parameters α1 and α2 has been described as an example. However, the normal range may be defined in N dimensions (for example, three or more dimensions). .

According to the second embodiment, the control device 1 that controls the autonomous driving vehicle 5 learns in advance the normal range defining parameter that defines the normal range based on at least the driving conditions and environmental conditions of the autonomous driving vehicle 5. Change or select multiple normal range specification parameters that define the normal range based on operating conditions or environmental conditions, and all the parameters related to control are included in the normal range specified by the changed or selected normal range specification parameters If it is determined that the operation state of the autonomous driving vehicle 5 is normal, the normal range is defined based on an approximation function that approximates the distribution of the normal range defining parameter values. As a result, the control device 1 can appropriately detect the abnormality of the operation state of the automatic driving vehicle 5 based only on the proper normal range defining parameter corresponding to the driving condition and the environmental condition of the automatic driving vehicle 5. The detection performance can be improved. Further, since the control device 1 uses only normal range defining parameters necessary for abnormality detection, it is possible to reduce the amount of arithmetic processing of the control device 1 and to optimize the calculation load.

As explained above, in the second embodiment,
(4) The normal operation determination unit 4A is configured to define the normal range defined by the normal range defining parameter based on a function that approximates the value of the normal range defining parameter.

If comprised in this way, the normal operation | movement judgment part 4 can judge appropriately whether the parameter regarding control of the autonomous driving vehicle 5 is normal based on the simple approximate function which prescribes | regulates a normal range.
<Third Embodiment>
Next, a control device according to a third embodiment of the present invention will be described.
FIG. 8 is a block diagram illustrating a functional configuration of a normal operation determination unit 4B according to the third embodiment. In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As shown in FIG. 8, the normal operation determination unit 4B clusters the normal range defined by the normal range defining parameter, and clusters the distribution areas of a plurality of parameter values related to the control of the controlled device (for example, the autonomous driving vehicle 5) ( It defines (learns) based on the result of division. Then, the normal operation determination unit 4B determines whether it is normal based on the following processes (1) and (2).
(1) If all the values of the parameters related to the control are within the normal range obtained by clustering the distribution ranges of the normal range defining parameters α1 and α2, the operation state of the automatic driving vehicle 5 is determined to be normal, and the value of the abnormality flag Is set to 0 (zero).
(2) In other cases (at least, when there are no parameter values related to control within the clustered normal range), the operation state of the autonomous driving vehicle 5 is determined to be abnormal, and the value of the abnormality flag is set to 1. Set to.
Note that the boundary of the normal range may be clarified using SVM (Support Vector Machine) or the like for each cluster. In FIG. 8, the case where the normal range is defined in two dimensions of the normal range defining parameters α1 and α2 has been described as an example, but the normal range may be defined in N dimensions (for example, three or more dimensions). .

According to the third embodiment, the control device 1 that controls the autonomous driving vehicle 5 learns in advance the normal range defining parameter that defines the normal range based on at least the driving conditions and environmental conditions of the autonomous driving vehicle 5. Change or select a plurality of normal range defining parameters that define the normal range based on operating conditions or environmental conditions, and all the parameters related to control are within the normal range defined by the changed or selected normal range defining parameters If it is included, it can be determined that the operation state of the autonomous driving vehicle 5 is normal, and the normal range is defined based on the result of clustering the distribution range of normal range defining parameter values. As a result, the control device 1 can appropriately detect the abnormality of the operation state of the automatic driving vehicle 5 based only on the proper normal range defining parameter corresponding to the driving condition and the environmental condition of the automatic driving vehicle 5. The detection performance can be improved. Further, since the control device 1 uses only normal range defining parameters necessary for abnormality detection, it is possible to reduce the amount of arithmetic processing of the control device 1 and to optimize the calculation load.

As explained above, in the third embodiment,
(5) The normal operation determining unit 4B is configured to define the normal range defined by the normal range defining parameter based on the result of clustering the values of the normal range defining parameter.

With this configuration, the normal range is set by clustering the distribution range of the normal range defining parameter, so that the normal range can be appropriately set without being affected by the distribution of the normal range defining parameter.

<Fourth embodiment>
Next, a control device according to a fourth embodiment of the present invention will be described.
FIG. 9 is a block diagram illustrating a functional configuration of the parameter learning unit 2D according to the fourth embodiment. In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As shown in FIG. 9, the parameter learning unit 2D performs a principal component analysis on the normal range defining parameter values related to normal control in the driving conditions and the environmental conditions transmitted from the autonomous driving vehicle 5, and performs the principal component analysis. As a result, the storage device 11 may store a dimension parameter for which the eigenvalue obtained is equal to or greater than a predetermined threshold.

According to the fourth embodiment, the control device that controls the autonomous driving vehicle 5 learns in advance the normal range defining parameter that defines the normal range based on at least the driving condition and the environmental condition of the autonomous driving vehicle 5, Change or select multiple normal range specification parameters that define the normal range based on operating conditions or environmental conditions, and all parameters related to control are included in the normal range specified by the changed or selected normal range specification parameters If so, it can be determined that the operating state of the autonomous driving vehicle 5 is normal. The normal range is defined by the upper limit value and the lower limit value of the normal range defining parameter for control, or is defined based on an approximation function that approximates the distribution of the normal range defining parameter value for control, or the normal range for control. It is defined based on the result of clustering the distribution region of the defined parameter value. In addition, normal range parameters that define the normal range are learned by performing a principal component analysis on the values of the normal range definition parameters related to normal control, and selecting a parameter whose eigenvalue obtained by the principal component analysis is a predetermined value or more. Use. Thereby, the control device can appropriately detect the abnormality of the operation state of the automatic driving vehicle 5 based only on the proper normal range defining parameter according to the driving condition and the environmental condition of the automatic driving vehicle 5. Performance can be improved. Further, since the control device uses only normal range defining parameters necessary for abnormality detection, it is possible to reduce the amount of arithmetic processing of the control device and optimize the calculation load.

As described above, in the fourth embodiment,
(6) Configuration in which the parameter learning unit 2D learns a normal range defining parameter of a dimension in which an eigenvalue obtained as a result of performing principal component analysis on a normal range defining parameter value that defines a normal range is equal to or greater than a predetermined threshold. It was.

With this configuration, the parameter learning unit 2D can accurately extract a parameter having a large contribution to the control of the autonomous driving vehicle 5 as a normal range defining parameter by principal component analysis.

<Fifth embodiment>
Next, a control device according to a fifth embodiment of the present invention will be described.
FIG. 10 is a block diagram illustrating a functional configuration of a parameter learning unit 2E according to the fifth embodiment. In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As shown in FIG. 10, the parameter learning unit 2E performs the Lasso regression on the normal range regulation parameter value regarding the normal control in the driving condition and the environmental condition transmitted from the autonomous driving vehicle 5, and the result of performing the Lasso regression. The remaining normal range defining parameters may be stored in the storage device 11.

According to the fifth embodiment, the control device that controls the autonomous driving vehicle 5 learns the normal range parameter that defines the normal range based on at least the driving condition and the environmental condition of the autonomous driving vehicle 5 in advance. Change or select multiple normal range specification parameters that define the normal range based on conditions or environmental conditions, and all the parameters related to control are included in the normal range specified by the changed or selected normal range specification parameter If so, it can be determined that the operating state of the autonomous driving vehicle 5 is normal. The normal range is defined by the upper limit value and the lower limit value of the normal range defining parameter for control, or is defined based on an approximation function that approximates the distribution of the normal range defining parameter value for control, or the normal range for control. It is defined based on the result of clustering the distribution region of the defined parameter value. Moreover, the learning of a predetermined range uses the normal range prescription parameter which remained as a result of performing Lasso regression. Thereby, the control device can appropriately detect the abnormality of the operation state of the automatic driving vehicle 5 based only on the proper normal range defining parameter according to the driving condition and the environmental condition of the automatic driving vehicle 5. Performance can be improved. Further, since the control device uses only normal range defining parameters necessary for abnormality detection, it is possible to reduce the amount of arithmetic processing of the control device and optimize the calculation load.

As explained above, in the fifth embodiment,
(7) The parameter learning unit 2E is configured to learn the normal range defining parameters of the remaining dimensions as a result of performing the Lasso regression process on the normal range defining parameter values that define the normal range.

With this configuration, the parameter learning unit 2E can efficiently extract a parameter having a large contribution to the control of the autonomous driving vehicle 5 as a normal range defining parameter by the Lasso regression process.

<Sixth Embodiment>
Next, a control device according to a sixth embodiment of the present invention will be described.
FIG. 11 is a block diagram illustrating a functional configuration of a control device 1F according to the sixth embodiment.
FIG. 12 is a block diagram when the control device 1F according to the sixth embodiment is applied to an autonomous driving vehicle.
FIG. 13 is a block diagram illustrating a functional configuration of a parameter learning unit 2F according to the sixth embodiment.
FIG. 14 is a block diagram illustrating a functional configuration of a parameter changing unit 3F according to the sixth embodiment.
FIG. 15 is a block diagram illustrating a functional configuration of a normal operation determination unit 4F according to the sixth embodiment.
In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As illustrated in FIG. 11, the control device 1F includes a parameter learning unit 2F that learns a normal range of parameters related to control based on operating conditions and environmental conditions transmitted from the controlled device, and a plurality of normal states that define the normal range. It is determined that the operating state of the controlled device is normal when the range defining parameter, the parameter changing unit 3F for changing or selecting the normal range, and all the parameters related to the control transmitted from the controlled device are included in the normal range. And a normal operation determination unit 4F.

As shown in FIG. 12, the control device 1F is connected to an autonomous driving vehicle 5 that is a controlled device, and from the autonomous driving vehicle 5, movement conditions (for example, vehicle speed, acceleration, etc.) and environmental conditions (for example, air temperature, Humidity) and parameters related to control (for example, target throttle opening, target fuel injection amount, target yaw rate, etc.) are transmitted to the control device 1F.

<Parameter learning unit>
As shown in FIG. 13, the parameter learning unit 2F determines whether or not the operation state of the automatic driving vehicle 5 is normal based on the driving conditions and environmental conditions transmitted from the automatic driving vehicle 5 that is the controlled device. The normal ranges of the parameter values (α1, α2 and β1, β2) are learned. Specifically, the following processes (1) and (2) are performed.
(1) In the driving conditions and environmental conditions transmitted from the autonomous driving vehicle 5, the difference e between the maximum value and the minimum value of each dimension of the parameters (α1, α2, and β1, β2) relating to normal control is predetermined. A parameter having a dimension equal to or greater than the threshold value K1 is extracted and stored in the storage device 11.
(2) The top of the normal space is defined by the upper limit value and the lower limit value of the normal range defining parameters (α1, α2, and β1, β2) in the driving conditions and environmental conditions transmitted from the autonomous driving vehicle 5, and stored in the storage device 11. Remember.
The operating conditions and the environmental conditions may be represented by numbers corresponding to the conditions. Note that FIG. 12 illustrates the case of the three operating conditions and environmental conditions A, B, and C, but the number of operating conditions and environmental conditions is not limited thereto.

<Parameter change part>
As shown in FIG. 14, the parameter changing unit 3 is the normal learned by the parameter learning unit 2 in accordance with the driving conditions and environmental conditions (represented by numbers corresponding to the respective conditions) transmitted from the autonomous driving vehicle 5. Change or select a predetermined parameter from a plurality of normal range defining parameters that define the range. Specifically, the following processes (1) and (2) are performed.
(1) Based on the driving conditions and environmental conditions transmitted from the autonomous driving vehicle 5, a plurality of normal range defining parameters that define the normal range are changed or selected.
(2) Change or select a normal space according to the driving conditions and environmental conditions transmitted from the autonomous driving vehicle 5.
In addition, in FIG. 14, although the case of three conditions, driving | running conditions and environmental conditions A, B, and C is illustrated and demonstrated, the conditions number of driving | running conditions and environmental conditions is not restricted to this.

<Normal operation determination unit>
As shown in FIG. 15, the normal operation determining unit 4 has a parameter value related to control transmitted from the autonomous driving vehicle 5 within the normal range defined by the normal range defining parameter changed or selected by the parameter changing unit 3. When all are present, it is determined that the operation state of the automatic driving vehicle 5 is normal, and when at least some of the parameter values are not within the normal range, it is determined that the operation state of the automatic driving vehicle 5 is abnormal. Specifically, the following processes (1) to (3) are performed.
(1) The normal range is changed or selected according to the driving condition and the environmental condition transmitted from the autonomous driving vehicle 5 together with the normal range defining parameter.
(2) When all the values of the parameters relating to the control are within the normal range defined by the normal range defining parameter, it is determined that the operation state of the automatic driving vehicle 5 is normal, and the value of the abnormality flag is set to 0 (zero). Set to.
(3) In other cases (at least, when there are no parameter values related to control within the normal range), the operation state of the autonomous driving vehicle 5 is determined to be abnormal, and the value of the abnormality flag is set to 1. To do.
FIG. 15 illustrates the case where the normal range is defined in two dimensions of normal range defining parameters α1 and α2, β1 and β2, or θ1 and θ2, respectively. The normal range may be defined as described above.

According to the sixth embodiment, the control device 1F that controls the autonomous driving vehicle 5 learns in advance the normal range defining parameter that defines the normal range based on at least the driving conditions and environmental conditions of the autonomous driving vehicle 5. When the normal range is changed or selected and a normal range is defined based on operating conditions or environmental conditions, and all parameters related to control are included in the changed or selected normal range, It can be determined that the operating state of the autonomous driving vehicle 5 is normal. The normal range is defined by an upper limit value and a lower limit value of parameters related to control. Thereby, the control device 1F can appropriately detect the abnormality of the operation state of the automatic driving vehicle 5 based only on the proper normal range defining parameter corresponding to the driving condition and the environmental condition of the automatic driving vehicle 5. The detection performance can be improved. In addition, since the control device 1F uses only normal range defining parameters necessary for abnormality detection, it is possible to reduce the amount of calculation processing of the control device 1F and optimize the calculation load.

As explained above, in the sixth embodiment,

(8) The parameter changing unit 3F is configured to change or select a normal range for determining whether or not the operation state of the autonomous driving vehicle 5 is normal.

If comprised in this way, since the control apparatus 1F can change and select the normal range suitable for the control according to the driving condition and environmental condition of the autonomous driving vehicle 5, it will appropriately judge the abnormal state based on the driving condition and the environmental condition. can do.

<Seventh embodiment>
Next, a control device according to a seventh embodiment of the present invention will be described.
FIG. 16 is a block diagram illustrating a functional configuration of a control device 1G according to the seventh embodiment. The seventh embodiment is different from the above-described embodiment in that the controlled device controlled by the control device 1G is the robot 6. In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As shown in FIG. 16, the control device 1G is connected to the robot 6, and the operating conditions (for example, the fulcrum positions of joints of the robot arms) transmitted from the robot 6 and the environmental conditions (for example, temperature, humidity, etc.) ) Is transmitted to the control device 1G. The parameter learning unit 2G learns a normal range defining parameter that defines a normal range in which the operation state of the robot 6 is normal based on the operation condition and the environmental condition transmitted from the robot 6. Then, the parameter changing unit 3G changes or selects a plurality of normal range defining parameters learned by the parameter learning unit 2G according to the operating conditions and the environmental conditions. When the normal operation determining unit 4G includes all of the parameters related to the control transmitted from the robot 6 within the normal range defined by the normal range defining parameter changed or selected by the parameter changing unit 3G, If it is determined that the operation state of the robot 6 is normal, the abnormality flag is set to 0 (zero), and at least some of the parameters relating to the control are not included, it is determined that the operation state of the robot 6 is abnormal, The abnormality flag is set to 1.

With this configuration, the control device 1G can detect the abnormality of the operation state of the robot 6 based only on the proper normal range defining parameter corresponding to the operation condition and the environmental condition of the robot 6, and the abnormality detection Performance can be improved. Further, since only the normal range defining parameters necessary for abnormality detection are used, the amount of calculation processing of the control device 1G can be reduced and the calculation load can be optimized.

<Eighth Embodiment>
Next, a control device according to an eighth embodiment of the present invention will be described.
FIG. 17 is a block diagram illustrating a functional configuration of a control device 1H according to the eighth embodiment. The eighth embodiment is different from the above-described embodiment in that the controlled device controlled by the control device 1H is the flying object 7 (for example, an unmanned drone). In addition, about the same structure and function as the control apparatus 1 concerning 1st Embodiment, the same code | symbol is attached | subjected and it demonstrates as needed.

As shown in FIG. 17, the control device 1H is connected to the flying object 7, and the operating conditions (for example, the moving speed and acceleration of the flying object) and environmental conditions (for example, air temperature and humidity) transmitted from the flying object 7 are transmitted. , Wind speed, etc.) are transmitted to the control device 1H. The parameter learning unit 2H learns a normal range defining parameter that defines a normal range in which the operating state of the flying object 7 is normal based on the driving conditions and environmental conditions transmitted from the flying object 7. Then, the parameter changing unit 3H changes or selects a plurality of normal range defining parameters learned by the parameter learning unit 2H according to the operating condition and the environmental condition. When the normal range of the normal range regulation parameter changed or selected by the parameter changing unit 3H includes all the parameters related to the control transmitted from the flying object 7, the normal operation determining unit 4H 7 is determined to be normal, the abnormality flag is set to 0 (zero), and if at least some of the control parameters are not included, it is determined that the operational state of the flying object 7 is abnormal, The abnormality flag is set to 1.

If comprised in this way, the control apparatus 1H can perform the abnormality detection of the operation | movement of the flying body 7 only based on the appropriate normal range prescription | regulation parameter according to the driving | running condition and the environmental condition of the flying body 7, and abnormality detection Performance can be improved. Moreover, since only the normal range defining parameters necessary for abnormality detection are used, the amount of calculation processing of the control device 1H can be reduced and the calculation load can be optimized.

In the above, an example of the embodiment of the present invention has been described. However, the present invention may combine all of the above-described embodiments, and may arbitrarily combine any two or more embodiments.

Further, the present invention is not limited to the one having all the configurations of the above-described embodiment, and a part of the configuration of the above-described embodiment is replaced with the configuration of another embodiment. In addition, the configuration of the above-described embodiment may be replaced with the configuration of another embodiment.

Further, a part of the configuration of the above-described embodiment may be added to, deleted from, or replaced with the configuration of another embodiment.

1: control device, 11: storage device, 12: CPU, 13: ROM, 14: RAM, 15: data bus, 16: input circuit, 17: input / output port, 18: output circuit, 2: parameter learning unit, 3 : Parameter changing unit, 4: normal operation determining unit, 5: vehicle, 6: robot, 7: flying object

Claims

A control device for controlling a controlled device,
All of the parameter values related to the control transmitted from the controlled device based on the operation state of the controlled device are included in the normal range defined by the plurality of normal range defining parameters related to the control of the controlled device. A normal operation determining unit that determines that the operating state of the controlled device is normal;
A parameter learning unit that learns the normal range defining parameter that defines the normal range based on an operating condition or an environmental condition of the controlled device;
A control device comprising: a parameter changing unit that changes or selects the normal range defining parameter based on an operating condition or an environmental condition of the controlled device during operation of the controlled device.
2. The control device according to claim 1, wherein the normal operation determining unit defines the normal range defined by the normal range defining parameter by an upper limit value and a lower limit value of the normal range defining parameter.
2. The control device according to claim 1, wherein the normal operation determining unit defines the normal range defined by the normal range defining parameter based on a function approximating a value of the normal range defining parameter.
The control device according to claim 1, wherein the normal operation determining unit defines the normal range defined by the normal range defining parameter based on a result of clustering values of the normal range defining parameter.
The parameter learning unit learns the normal range defining parameter of a dimension in which a difference between a maximum value and a minimum value in each dimension of the normal range defining parameter value that defines the normal range is a predetermined threshold value or more. The control device described.
The parameter learning unit learns a normal range defining parameter having a dimension in which an eigenvalue obtained as a result of performing a principal component analysis on the normal range defining parameter value defining the normal range is a predetermined threshold or more. The control device according to 1.
The control device according to claim 1, wherein the parameter learning unit learns a normal range defining parameter of a dimension remaining as a result of performing a Lasso regression process on the normal range defining parameter value that defines the normal range.
The control device according to claim 1, wherein the parameter changing unit changes or selects the normal range for determining whether or not the operation state of the controlled device is normal.
The controlled device is an autonomous driving vehicle,
The control device according to claim 1, wherein the control device controls the autonomous driving vehicle.
The controlled device is a robot;
The control device according to claim 1, wherein the control device controls the robot.
The controlled device is a flying object;
The control device according to claim 1, wherein the control device controls the flying object.
A control method for controlling a controlled device, comprising:
All of the parameter values related to the control transmitted from the controlled device based on the operation state of the controlled device are included in the normal range defined by the plurality of normal range defining parameters related to the control of the controlled device. A normal operation determining step for determining that the operating state of the controlled device is normal;
A parameter learning step for learning the normal range defining parameter for defining the normal range based on an operating condition or an environmental condition of the controlled device;
And a parameter changing step of changing or selecting the normal range defining parameter based on an operating condition or an environmental condition of the controlled device during operation of the controlled device.