WO2018008503A1 - Control device - Google Patents

Abstract

The present invention addresses the problem of controlling a behavior by using wide-range, global information. The present invention provides a control device constituted of a server and a plurality of terminals, characterized by being provided with: an occurrence probability prediction means for predicting occurrence probabilities of a plurality of different results that can occur in the future on the basis of past information; a priority calculation means for calculating priorities of a plurality of control targets on the basis of at least the occurrence probabilities of the plurality of different results, information detected in the terminals, and a parameter calculated in the terminals; and a control signal calculation means for calculating a control signal for determining an operation of the terminals on the basis of at least the control target with the highest priority.

Description

Control device

The present invention relates to a control device, and more particularly to a control device including a server and a plurality of terminals.

There is Patent Document 1 as background art in this technical field. This document includes "an input signal detection means for detecting an external input signal from the outside, an evaluation means for evaluating an external input signal detected by the input signal detection means, and an evaluation result by the evaluation means as action content information. A robot apparatus comprising: an association means for associating with an action control means for controlling an action based on the action content information based on the evaluation associated with the association means; (See [Claim 1]).

JP 2002-269530 A

However, the above-mentioned prior art (Patent Document 1) controls behavior using only information that can be detected by a single robot (terminal), and controls behavior using a wider range of global information. Is not something to do.

The present invention has been made in view of the above points, and an object of the present invention is to provide a control device that enables behavior control according to the situation of each terminal while using a wider range of global information. Is to provide.

In order to solve the above problems, for example, the configuration described in the claims is adopted. The present invention includes a plurality of means for solving the above-described problems. For example, the control device includes a server and a plurality of terminals, and based on past information, occurrence probability of a plurality of different results that may occur in the future. Based on at least the occurrence probability of the plurality of different results, the information detected at the terminal, and the parameter calculated at the terminal. Priority calculating means for calculating, and at least control signal calculating means for calculating a control signal for determining the operation of the terminal based on the control target having the highest priority.

In addition, for example, the parameter calculated in the terminal is a parameter that weights at least the occurrence probability of the plurality of different results and the information detected in the terminal, and the priority calculation means includes the weight It is characterized in that the priority of a plurality of control targets is calculated based on the value subjected to the attaching process.

Further, for example, the occurrence probability prediction means includes a parameter value update means for updating a parameter value of the occurrence probability prediction means based on newly obtained information.

Further, for example, it is characterized by comprising terminal parameter updating means for updating a parameter calculated in the terminal for each terminal.

Further, for example, the information detected in the terminal is at least a value based on a sensor output value installed in the terminal.

Further, for example, the information detected in the terminal is a value that is uniquely learned based on at least a sensor output value installed in the terminal.

Further, for example, the occurrence probability predicting means is at least a Bayesian network or a neural network.

Further, for example, the priority calculation means is at least logistic regression or reinforcement learning.

Further, for example, the occurrence probability prediction means includes a parameter value update means for updating the parameter value of the occurrence probability prediction means at least by Bayesian update or probabilistic gradient descent based on newly obtained information. Features.

Further, for example, the control target of the control device is a moving body such as an autonomous driving vehicle, and the plurality of control targets are moving paths of the moving body.

Further, for example, the control target of the control device is an internal combustion engine, and the plurality of control targets are operating conditions of the internal combustion engine.

Further, for example, the control target of the control device is a robot, and the plurality of control targets are tasks performed by the robot.

Further, for example, the occurrence probability prediction means includes a parameter value update means for updating the parameter value of the occurrence probability prediction means based on information newly obtained from the terminal.

Further, for example, a plurality of different results that may occur in the future are set as the plurality of control targets.

Further, for example, the current information is used as input information of the occurrence probability prediction means.

Further, for example, the probability of occurrence of a plurality of different results that may occur in the future is at least the probability that the mobile object will arrive at the target value on time, or the probability that the energy efficiency / fuel efficiency of the mobile object will fall within a predetermined range, or Probability that the motor / engine torque can exceed the predetermined value, or the probability that the fuel efficiency improvement rate of the prime mover / engine is greater than the predetermined value, or the probability that the fuel efficiency deterioration rate of the prime mover / engine exceeds the predetermined value, or It is characterized by the probability that a task will be completed within a predetermined time, or the probability that the risk of the robot will increase.

According to the present invention, the server predicts the probability of occurrence of a plurality of different results that may occur in the future using a wider range of global information, and the terminal detects the occurrence probability of the plurality of different results at the terminal. Since the behavior of the terminal is controlled on the basis of the information, the behavior control according to the situation of each terminal can be performed while using a wider range of global information.

Further features related to the present invention will become apparent from the description of the present specification and the accompanying drawings. Further, problems, configurations, and effects other than those described above will be clarified by the following description of embodiments.

FIG. 4 is an overall view of a control device in Examples 1 to 4. FIG. 6 is a system diagram of the server side on which the control device in Examples 1 to 4 and 6 operates. FIG. 6 is a system diagram of the terminal side on which the control device in Examples 1 to 4 and 6 operates. The figure which showed the process of the means to predict the generation probability of the several different result which may occur in the future based on the past information in the server in Example 1, 4, 6. FIG. The figure which showed the process in the terminal in Example 1, 4, 5. FIG. The figure which showed the process of the means to calculate the priority of the some control target in the terminal in Example 1, 5. FIG. The figure which showed the process of the means which calculates the control signal in the terminal in Example 1, 4, 5, 6. The figure which showed the process of the means to estimate the generation probability of the several different result which may occur in the future based on the past information in the server in Example 2. FIG. The figure which showed the process in the terminal in Example 2. FIG. The figure which showed the process of the means to calculate the priority of the some control target in the terminal in Example 2. FIG. The figure which showed the process of the means which calculates the control signal in the terminal in Example 2. FIG. The figure which showed the process of the means to estimate the generation probability of the several different result which may occur in the future based on the past information in the server in Example 3. FIG. The figure which showed the process in the terminal in Example 3. FIG. The figure which showed the process of the means to calculate the priority of the some control target in the terminal in Example 3. FIG. The figure which showed the process of the means to calculate the control signal in the terminal in Example 3. FIG. The figure which showed the process of the means to calculate the priority of the some control target in the terminal in Example 4. FIG. FIG. 10 is an overall view of a control device according to a fifth embodiment. FIG. 10 is a system diagram of the server side on which the control device according to the fifth embodiment operates. The system figure by the side of the terminal where the control apparatus in Example 5 operate | moves. The figure which showed the process of the means to predict the production | generation probability of the several different result which may occur in the future, and the means to update a parameter value based on the past information in the server in Example 5. FIG. FIG. 10 is an overall view of a control device in Embodiment 6. The system figure by the side of the terminal where the control apparatus in Example 6 operate | moves. The figure which showed the process of the means to calculate the priority of the several control target in the terminal in Example 6. FIG. The figure which showed the process of the means to update the parameter in the terminal in Example 6 for every terminal.

Hereinafter, examples will be described with reference to the drawings.
[Example 1]

In the present embodiment, an occurrence probability predicting means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information, and at least the occurrence probabilities of the plurality of different results are detected at the terminal. Priority calculating means for calculating the priority of a plurality of control targets based on information and parameters calculated in the terminal; and at least the operation of the terminal based on the control target having the highest priority. A mode provided with control signal calculation means for calculating a control signal for determination will be described.

In particular, the controlled object is a moving body that can be moved unattended, such as an autonomous driving vehicle.

Also, the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network.

Also, the current information is used as the input information of the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on the past information.

Also, information detected at the terminal is a sensor output value installed at the terminal.

Further, the parameter calculated in the terminal weights at least the occurrence probability of the plurality of different results and the information detected in the terminal.

FIG. 1 is a diagram showing the entire control device. The server 1 includes an occurrence probability predicting unit 2 that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information, and calculates the occurrence probability of a plurality of different results that may occur in the future. Probabilities of occurrence of a plurality of different results that may occur in the future are sent to the plurality of terminals 3 by wired communication or wireless communication. The terminal 3 includes priority calculation means 4 that calculates the priority of a plurality of control targets based on the occurrence probability of a plurality of different results, information detected at the terminal, and parameters calculated at the terminal. And calculate the priority of the control target. Then, based on the control target with the highest priority, the control signal is calculated by the control signal calculation means 5 for calculating the control signal for determining the operation of the terminal.

FIG. 2 is a system diagram of the server 1. The server 1 is provided with an input circuit 16 for processing an external signal. Examples of the signal from the outside here include information from outside the server such as weather, temperature, road information, information detected by the sensor of the terminal 3, and the like. An external signal passes through the input circuit and becomes an input signal and is sent to the input / output port 17. The input information sent to the input / output port 17 is written into the RAM 14 through the data bus 15. Alternatively, it is stored in the storage device 11. Various processes are written in the ROM 13 or the storage device 11 and executed by the CPU 12. At that time, the value written in the RAM 14 or the storage device 11 is used as appropriate for calculation. Of the calculation results, information (value) to be sent to the outside is sent to the input / output port 17 through the data bus 15 and sent to the output circuit 18 as an output signal. The signal is output from the output circuit 18 to the outside. The signal to the outside here is an occurrence probability of a plurality of different results that may occur in the future, and is sent to a plurality of terminals 3.

As described above with reference to FIG. 1, the server 1 calculates the probability of occurrence of a plurality of different results that may occur in the future. Using the input signal written in the RAM 14 or the storage device 11 via the input circuit 16 and the input / output port 17, the CPU 13 is used to perform later-described processing stored in the ROM 13 or the storage device 11 to calculate the output signal. To do. As described above, the output signal is output (transmitted to the terminal 3) as a signal to the outside (probability of occurrence of a plurality of different results that may occur in the future) via the RAM 14, the input / output port 17, and the output circuit 18.

FIG. 3 is a system diagram of the terminal 3. The terminal 3 is provided with an input circuit 26 for processing an external signal. The signal from the outside here may be, for example, a sensor signal installed in the terminal 3. These external signals are input to the input / output port 27 via the input circuit 26 as input signals. Further, from the server 1, information on the probability of occurrence of a plurality of different results that may occur in the future is input to the input / output port 27. Each input information sent to the input / output port 27 is written into the RAM 24 through the data bus 25. Alternatively, it is stored in the storage device 21. Various processes are written in the ROM 23 or the storage device 21 and are executed by the CPU 22. At that time, the value written in the RAM 24 or the storage device 21 is used as appropriate for calculation. Of the calculation results, information (value) to be sent to the outside is sent to the input / output port 27 through the data bus 25 and sent to the output circuit 28 as an output signal. The signal is output from the output circuit 28 to the outside. The signal to the outside here refers to an actuator signal or the like for causing the control target to make a desired movement.

As described above with reference to FIG. 1, in the terminal 3, the priority calculation means 4 calculates the priority of the control target, and the control signal calculation means 5 calculates the control signal. An input signal written in the RAM 24 or the storage device 21 via the input circuit 26 and the input / output port 27 and a signal from the server 1 (probability of occurrence of a plurality of different results that may occur in the future) are stored in the ROM 23 or the storage device 21. The stored processing to be described later is performed using the CPU 22 to calculate the output signal. As described above, the output signal passes through the RAM 24, the input / output port 27, and the output circuit 28, and is output as an external signal (control signal). Details of each process will be described below.

<Occurrence probability predicting means 31 (FIG. 4) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information>
In this process, the probability of occurrence of a plurality of different results that may occur in the future is calculated based on past information. Specifically, it is shown in FIG.
-As input information, current information such as weather, temperature, road information, and time are used.
Use a Bayesian network to calculate the probability of the next four possible future outcomes.

(1) Probability of arriving at the target value on time when route A is selected (2) Probability of arriving at the target value on time when route B is selected (3) When selecting route A The probability that the fuel consumption deterioration rate will be 10% or more (4) The probability that the fuel consumption deterioration rate will be 10% or more when route B is selected

In the Bayesian network, there is information representing the relationship between past weather, temperature, road information, time information and the above four occurrence probabilities. That is, the probability of occurrence of the above-mentioned four different results that may occur in the future is predicted based on the past information of weather, temperature, road information, and time.

The details of the Bayesian network have been described in many documents and books, and will not be described in detail here.

FIG. 5 shows an overall view of processing in the terminal 33.
The priority calculating means 34 for calculating the priorities of a plurality of control targets calculates the priority of the movement route that is the priority of the control target.
The control signal calculating means 35 for calculating the control signal calculates a control signal for traveling on the determined moving route, and controls the moving body 32 that is a control target.

Next, details of the priority calculation means 34 and the control signal calculation means 35 will be described.

<Priority calculation means for calculating priorities of a plurality of control targets (FIG. 6)>
In this process, priorities of a plurality of control targets are calculated. Specifically, it is shown in FIG.
The four probabilities calculated by the occurrence probability prediction means 31 (see FIG. 4) and the information detected at the terminal are, for example, the following, which are values that can be detected by a sensor or input device installed in the terminal.
・ Used years (total mileage)
・ Today's total mileage ・ Driver characteristics (gender / age / driving history / fatigue)
・ The location of the vehicle and the remaining fuel.
In addition, since the detection method of each said information is described in many literatures and books, it does not elaborate here.

The priority calculating means 36 for calculating the priority for selecting the route A multiplies the above four probabilities and the information detected by the terminal 33 by weighting factors a_11, a_12,. The priority for selecting is calculated.

The priority calculation means 37 for calculating the priority for selecting the route B multiplies the above four probabilities and the information detected by the terminal by weighting factors a_21, a_22,. Calculate the priority to choose.

<Control signal calculation means (FIG. 7)>
In this process, a control signal for traveling on the determined movement route is calculated. Specifically, it is shown in FIG.
The calculation means 38 for determining the route with the highest priority as the moving route determines the route with the highest priority among the priority for selecting the route A and the priority for selecting the route B as the moving route. When the priority values are the same, the route A or the route B may be selected. Decide in advance.
A calculation means 39 for calculating a control signal for traveling on the determined movement route calculates a control signal for traveling on the determined movement route.

This means will not be described in detail because there are documents and materials. It is conceivable to adjust the drive signal to the prime mover (motor, engine), the drive signal to the rotating device (electronic control steering), etc. for controlling the speed and rotation angle of the moving body while inquiring the current position.

The occurrence probability prediction means (see FIG. 4) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network, but a similar function can be realized as a neural network.

Further, in the priority calculation means (see FIG. 6) for calculating the priorities of a plurality of control targets, the priority of the route to be selected is calculated, but a configuration in which the priority of the fuel consumption deterioration rate is also considered.

[Example 2]
In the present embodiment, an occurrence probability predicting means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information, and at least the occurrence probabilities of the plurality of different results are detected at the terminal. Priority calculating means for calculating the priority of a plurality of control targets based on information and parameters calculated in the terminal; and at least the operation of the terminal based on the control target having the highest priority. A mode provided with control signal calculation means for calculating a control signal for determination will be described.

In particular, an object to be controlled is an engine (internal combustion engine) used in, for example, a motor for an automobile.
Moreover, the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network.
In addition, the current information is used as input information of an occurrence probability prediction unit that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information.
Moreover, the information detected in the terminal is a sensor output value installed in the terminal.
The parameter calculated in the terminal weights at least the occurrence probability of the plurality of different results and the information detected in the terminal.

FIG. 1 is a diagram showing the entire control apparatus, which is the same as the first embodiment and will not be described in detail.
FIG. 2 is a system diagram of the server 1, which is the same as that of the first embodiment and will not be described in detail.
FIG. 3 is a system diagram of the terminal 3, which is the same as in the first embodiment, and will not be described in detail.

The details of each process are described below.

<Occurrence probability predicting means 41 (FIG. 8) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information>
In this process, the probability of occurrence of a plurality of different results that may occur in the future is calculated based on past information. Specifically, it is shown in FIG.
-As the input information, the current information such as the weather within a radius of 10 km, the average temperature within a radius of 10 km, the average humidity within a radius of 10 km, and road information are used.
Use a Bayesian network to calculate the probability of the next four possible future outcomes.

(1) Probability that the maximum torque can be produced (2) Probability that the fuel efficiency improvement rate will be 10% or more (3) Probability that the fuel efficiency deterioration rate will be 10% or more (4) Probability that the service life of the engine will be shortened by 1 year or more

In the Bayesian network, there is information representing the relationship between each of the above four occurrence probabilities, the weather within the past 10km radius, the average temperature within the 10km radius, the average humidity within the 10km radius, and road information. In other words, the probability of occurrence of the above-mentioned four different results that can occur in the future is predicted based on the information of the weather within the radius of 10 km, the average temperature within the radius of 10 km, the average humidity within the radius of 10 km, and the road information.

The details of the Bayesian network have been described in many documents and books, and will not be described in detail here.

FIG. 9 shows an overall view of processing at the terminal 46.
The priority calculation means 47 for calculating the priority of a plurality of control targets calculates the priority of the engine operation mode (operating condition of the internal combustion engine) that is the priority of the control target.
The control signal calculation means 48 for calculating the control signal calculates a control signal for setting the determined engine operation mode, and controls the engine 42 to be controlled.

Next, details of the priority calculation means 47 and the control signal calculation means 48 will be described.

<Priority calculating means for calculating priorities of a plurality of control targets (FIG. 10)>
In this process, priorities of a plurality of control targets are calculated. Specifically, it is shown in FIG.
The four probabilities calculated by the occurrence probability predicting means 41 (see FIG. 8) and the information detected by the terminal 46 are, for example, the following values that can be detected by a sensor or an input device installed in the terminal 46. .
・ Used years (total mileage)
・ Temperature, humidity, remaining fuel, etc.

In addition, since the detection method of each said information is described in many literatures and books, it does not elaborate here.

In the priority calculation means 49 for calculating the priority for selecting the power mode, the above four probabilities and the information detected by the terminal 46 are multiplied by weighting factors a_11, a_12,. The priority for selecting is calculated.

The priority calculation means 50 for calculating the priority for selecting the fuel consumption mode multiplies the above four probabilities and the information detected at the terminal by weighting factors a_21, a_22,. Calculate the priority to choose.

The priority calculation means 51 for calculating the priority for selecting the intermediate mode multiplies the above four probabilities and the information detected by the terminal by weighting factors a_31, a_32,. Calculate the priority to choose.

<Control signal calculating means for calculating the control signal (FIG. 11)>
In this process, a control signal for setting the determined operation mode is calculated. Specifically, it is shown in FIG.
The calculation means 52 that determines the mode with the highest priority as the operation mode determines the mode with the highest priority among the power mode, the fuel consumption mode, and the intermediate mode as the operation mode.
The control means 53 for calculating the control signal for setting the determined operation mode calculates the control signal for setting the determined operation mode.

This means will not be described in detail because there are documents and materials. It is conceivable to set a desired operation mode according to the air-fuel ratio based on the fuel injection amount and the air amount, and the ignition timing.

The occurrence probability prediction means (see FIG. 8) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network, but a similar function can be realized as a neural network.

Further, in the priority calculation means (see FIG. 10) for calculating the priority of the plurality of control targets, the priority of the engine operation mode is calculated. Based on the past information, a plurality of different results that may occur in the future are calculated. An operation mode for generating the maximum torque, which is a result calculated by the occurrence probability predicting means 41 (FIG. 8) for predicting the occurrence probability.
According to the present embodiment, power, fuel consumption, life, energy efficiency, etc. can be set according to the operation history of each engine. Therefore, engine control according to the user's preference can be performed.

[Example 3]
In the present embodiment, an occurrence probability predicting means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information, and at least the occurrence probabilities of the plurality of different results are detected at the terminal. Priority calculating means for calculating the priority of a plurality of control targets based on information and parameters calculated in the terminal; and at least the operation of the terminal based on the control target having the highest priority. A mode provided with control signal calculation means for calculating a control signal for determination will be described.

In particular, the controlled object is a robot.
Moreover, the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network.
In addition, the current information is used as input information of an occurrence probability prediction unit that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information.
Moreover, the information detected in the terminal is a sensor output value installed in the terminal.
The parameter calculated in the terminal weights at least the occurrence probability of the plurality of different results and the information detected in the terminal.

FIG. 1 is a diagram showing the entire control apparatus, which is the same as the first embodiment and will not be described in detail.
FIG. 2 is a system diagram of the server 1, which is the same as that of the first embodiment and will not be described in detail.
FIG. 3 is a system diagram of the terminal 3, which is the same as in the first embodiment, and will not be described in detail.

The details of each process are described below.

<Occurrence probability predicting means 61 (FIG. 12) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information>
In this process, the probability of occurrence of a plurality of different results that may occur in the future is calculated based on past information. Specifically, it is shown in FIG.
-As input information, the number of active robots, the number of stopped robots, the number of all packages to be moved, and the total travel distance of packages are used as input information.
Use a Bayesian network to calculate the probability of the next four possible future outcomes.

(1) Probability of completing a task within a predetermined time when in speed mode (2) Probability of completing a task within a predetermined time when in careful mode (3) Probability of increasing risk when in speed mode (4) Probability of increasing risk when using careful mode

Inside the Bayesian network, there is information representing the relationship between the above four occurrence probabilities and the information on the number of robots that have been operating in the past, the number of robots that have stopped, the number of all packages to be moved, and the total movement distance of packages. . That is, the probability of occurrence of the above-mentioned four different results that can occur in the future is predicted based on the information on the number of operating robots, the number of robots that are stopped, the number of all packages to be moved, and the total movement distance of the packages. .

The details of the Bayesian network have been described in many documents and books, and will not be described in detail here.

FIG. 13 shows an overall view of processing at the terminal 63.
The priority calculation means 64 for calculating the priority of a plurality of control targets calculates the priority of the robot operation mode, which is the priority of the control target.
The control signal calculation means 65 for calculating the control signal calculates a control signal for setting the determined robot operation mode, and controls the robot 62 to be controlled.

Next, details of the priority calculation means 64 and the control signal calculation means 65 will be described.

<Priority calculating means for calculating priorities of a plurality of control targets (FIG. 14)>
In this process, priorities of a plurality of control targets are calculated. Specifically, it is shown in FIG.
The four probabilities calculated by the occurrence probability predicting means 61 (see FIG. 12) and information detected by the terminal 63 are, for example, the following, which are values that can be detected by a sensor or input device installed in the terminal 63. .
・ Operating years, temperature, humidity, remaining battery power, presence of obstacles, presence of people, etc.

In addition, since the detection method of each said information is described in many literatures and books, it does not elaborate here.

The priority calculating means 66 for calculating the priority for selecting the speed mode multiplies the above four probabilities and the information detected by the terminal 63 by weighting factors a_11, a_12,. The priority for selecting is calculated.

The priority calculating means 67 for calculating the priority for selecting the careful mode multiplies the above four probabilities and the information detected by the terminal 63 by weighting factors a_21, a_22,. The priority for selecting is calculated.

The priority calculation means 68 for calculating the priority for selecting the intermediate mode multiplies the above four probabilities and the information detected by the terminal 63 by weighting factors a_31, a_32,. The priority for selecting is calculated.

<Control signal means for calculating the control signal (FIG. 15)>
In this process, a control signal for setting the determined operation mode is calculated. Specifically, it is shown in FIG.
The calculation unit 69 that determines the mode with the highest priority as the operation mode determines the mode with the highest priority among the speed mode, the cautious mode, and the intermediate mode as the operation mode.
The control means 70 for calculating the control signal for setting the determined operation mode calculates the control signal for setting the determined operation mode.

This means will not be described in detail because there are documents and materials. A drive signal to a motor for moving the robot's arms and legs can be considered.

The occurrence probability prediction means (see FIG. 12) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network, but a similar function can be realized as a neural network.

Moreover, in the priority calculation means (see FIG. 14) for calculating the priority of the plurality of control targets, the priority of the operation mode of the robot is calculated. Based on past information, a plurality of different results that may occur in the future are calculated. A configuration is also possible in which the task is terminated within a predetermined time, which is a result calculated by the occurrence probability predicting means 61 (FIG. 12) for predicting the occurrence probability.
According to the present embodiment, it is possible to select whether the production robot or the baggage handling robot emphasizes speed or safety, or whether the risk is high or low.

[Example 4]
In the present embodiment, an occurrence probability predicting means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information, and at least the occurrence probabilities of the plurality of different results are detected at the terminal. Priority calculating means for calculating the priority of a plurality of control targets based on information, a parameter calculated in the terminal and a value uniquely learned based on a sensor output value installed in the terminal; At least, a mode including control signal calculation means for calculating a control signal for determining the operation of the terminal based on the control target having the highest priority will be described.

In particular, it is a moving body that can be moved unattended, represented by an autonomous vehicle.
Moreover, the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network.
In addition, the current information is used as input information of an occurrence probability prediction unit that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information.
The information detected in the terminal is a value that is uniquely learned based on the sensor output value installed in the terminal and the sensor output value installed in the terminal.
The parameter calculated in the terminal weights at least the occurrence probability of the plurality of different results and the information detected in the terminal.

FIG. 1 is a diagram showing the entire control apparatus, which is the same as the first embodiment and will not be described in detail.
FIG. 2 is a system diagram of the server 1, which is the same as that of the first embodiment and will not be described in detail.
FIG. 3 is a system diagram of the terminal 3, which is the same as in the first embodiment, and will not be described in detail.

The details of each process are described below.

<Occurrence probability predicting means 31 (FIG. 4) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information>
In this process, the probability of occurrence of a plurality of different results that may occur in the future is calculated based on past information. Specifically, although shown in FIG. 4, it is the same as that of the first embodiment, and will not be described in detail.
FIG. 5 shows an overall view of the processing in the terminal 33, which is the same as that of the first embodiment and will not be described in detail.

Next, details of the priority calculation means 34 and the control signal calculation means 35 will be described.

<Priority calculating means for calculating priorities of a plurality of control targets (FIG. 16)>
In this process, priorities of a plurality of control targets are calculated. Specifically, it is shown in FIG.
The four probabilities calculated by the occurrence probability predicting means 31 (see FIG. 4) and information detected at the terminal are, for example, the following, which are values that can be detected by a sensor or input device installed in the terminal 33.
・ Used years (total mileage)
・ Today's total mileage ・ Driver characteristics (gender / age / driving history / fatigue)
・ The location of the vehicle and the remaining fuel.

Furthermore, as a value that is uniquely learned based on the sensor output value of the sensor installed in the terminal 33,
• Use user characteristics.
The user characteristic is a value that is uniquely learned based on the sensor output value of the sensor installed in the terminal. For example, it is conceivable to learn the frequency of selecting the route A (for example, a large road) and the frequency of selecting the route B (for example, a back road) from the travel history information. Users can also choose their preferences directly.

In the priority calculation means 71 for calculating the priority for selecting the route A, the above four probabilities, the information detected in the terminal, and the value learned independently based on the sensor output value installed in the terminal, The priority for selecting the route A is calculated by multiplying the weight coefficients a_11, a_12,.

In the priority calculation means 72 for calculating the priority for selecting the route B, the above four probabilities, the information detected in the terminal, and the value uniquely learned based on the sensor output value installed in the terminal, The priority for selecting the route B is calculated by multiplying the weight coefficients a_21, a_22,.

<Control Signal Calculation Means for Calculation of Control Signal (FIG. 7)>
In this process, a control signal for traveling on the determined movement route is calculated. Specifically, although shown in FIG. 7, it is the same as that of the first embodiment, and therefore will not be described in detail.

Although the occurrence probability prediction means 31 (see FIG. 4) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network, a similar function can also be realized as a neural network. .
In addition, in the priority calculation means (see FIG. 16) that calculates the priority of a plurality of control targets, the priority of the route to be selected is calculated, but a configuration in which the priority of the fuel consumption deterioration rate is also considered.

[Example 5]
In the present embodiment, an occurrence probability predicting means for predicting an occurrence probability of a plurality of different results that may occur in the future based on past information, and a plurality of events that may occur in the future from past information based on newly obtained information. Parameter value updating means for updating parameter values of means used for predicting the occurrence probability of different results, occurrence probability of the plurality of different results, information detected in the terminal, and parameters calculated in the terminal A priority calculation means for calculating the priority of a plurality of control targets, and a control signal calculation for calculating a control signal for determining the operation of the terminal based on at least the control target having the highest priority It shows about the form provided with the means.

In particular, it is a moving body that can be moved unattended, represented by an autonomous vehicle.
Moreover, the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network.
Further, the means for updating the parameter value of the occurrence probability prediction means used for predicting the occurrence probability of a plurality of different results that may occur in the future from the past information is based on newly obtained information such as information obtained from the terminal. The parameter update means is Bayesian update.
In addition, the current information is used as input information of an occurrence probability prediction unit that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information.
The information detected in the terminal is a value that is uniquely learned based on the sensor output value installed in the terminal and the sensor output value installed in the terminal.
The parameter calculated in the terminal weights at least the occurrence probability of the plurality of different results and the information detected in the terminal.

FIG. 17 is a diagram showing the entire control device. The server 81 includes an occurrence probability prediction unit 82 that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information, and a parameter value update unit 83 that updates a parameter value. Calculate the probability of occurrence of different results. Probabilities of occurrence of a plurality of different results that may occur in the future are sent to the plurality of terminals 84 by wired communication or wireless communication. The terminal 84 has priority calculation means 85 that calculates the priority of a plurality of control targets based on the occurrence probability of a plurality of different results, information detected at the terminal, and parameters calculated at the terminal 84. It is provided and calculates the priority of the control target. Then, based on the control target with the highest priority, the control signal is calculated by the control signal calculation means 86 for calculating the control signal for determining the operation of the terminal 84. Also, information detected by the terminal 84 is sent from the terminal 84 to the server 81 by wired communication or wireless communication.

FIG. 18 is a system diagram of the server 81, and information from the terminal 84 is input to the input / output port 17. Since other than that is the same as Example 1, it does not elaborate in detail.
FIG. 19 is a system diagram of the terminal 84, and information from the terminal 84 is output from the input / output port 27. Since it is the same in Example 1, it does not elaborate.

The details of each process are described below.

<Occurrence probability predicting means 87 for predicting the probability of occurrence of a plurality of different results that may occur in the future based on past information, and parameter value updating means 88 for updating parameter values (FIG. 20)>
In this process, the occurrence probability of a plurality of different results that may occur in the future is calculated based on past information, and the parameter value is updated. Specifically, it is shown in FIG.
-As input information, current information such as weather, temperature, road information, and time are used.
Use a Bayesian network to calculate the probability of the next four possible future outcomes.
(1) Probability of arriving at the target value on time when route A is selected (2) Probability of arriving at the target value on time when route B is selected (3) When selecting route A The probability that the fuel consumption deterioration rate will be 10% or more (4) The probability that the fuel consumption deterioration rate will be 10% or more when route B is selected

Within the Bayesian network, there is information representing the relationship between the past weather, temperature, road information, time information and the above four occurrence probabilities. That is, the probability of occurrence of the above-mentioned four different results that may occur in the future is predicted based on the past information of weather, temperature, road information, and time.
The Bayesian network parameter value updating means 88 updates the Bayesian network parameter values using a Bayesian updater using the current information such as weather, temperature, road information, time, and information from the terminal.

The information from the terminal 84 is, for example, a control result at the terminal 84,
-Arrival time when route A is selected-Arrival time when route B is selected-Whether route A is selected, is the fuel consumption deterioration rate 10% or higher?
・ Whether route B is selected, has the fuel consumption deterioration rate been 10% or more?
Can be considered. That is, the Bayesian network is updated based on the latest information so that a more accurate prediction probability can be obtained.

Note that the details of the Bayesian network and the Bayes renewal device are described in many documents and books, so they will not be described in detail here.

FIG. 5 shows an overall view of processing in the terminal, but since it is the same as in the first embodiment, it will not be described in detail.

<Priority calculation means for calculating priorities of a plurality of control targets (FIG. 6)>
In this process, priorities of a plurality of control targets are calculated. Specifically, it is shown in FIG. 6, but it is the same as that of the first embodiment, and therefore will not be described in detail.

<Control Signal Calculation Means for Calculation of Control Signal (FIG. 7)>
In this process, a control signal for traveling on the determined movement route is calculated. Specifically, although shown in FIG. 7, it is the same as that of the first embodiment, and therefore will not be described in detail.

Although the occurrence probability predicting means (see FIG. 20) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network, a similar function can be realized as a neural network. In this case, it is preferable to use a stochastic gradient descent method or a back propagation method to update the parameters.
In addition, in the priority calculation means (see FIG. 6) for calculating the priorities of a plurality of control targets, the priority of the route to be selected is calculated.

[Example 6]
In the present embodiment, an occurrence probability predicting means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information, and at least the occurrence probabilities of the plurality of different results are detected at the terminal. Priority calculating means for calculating the priority of a plurality of control targets based on information and parameters calculated in the terminal; and at least the operation of the terminal based on the control target having the highest priority. A mode signal control means for calculating a control signal for determination and a terminal parameter update means for updating a parameter calculated in the terminal for each terminal will be described.

In particular, the controlled object is a moving body that can be moved unattended, such as an autonomous driving vehicle.
Moreover, the occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network.
In addition, the current information is used as input information of an occurrence probability prediction unit that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information.
Moreover, the information detected in the terminal is a sensor output value installed in the terminal.
The parameter calculated in the terminal weights at least the occurrence probability of the plurality of different results and the information detected in the terminal.
Moreover, the function which updates the parameter calculated in the said terminal for every terminal is linear regression and logistic regression.

FIG. 21 is a diagram showing the entire control device. The server 1 includes an occurrence probability predicting unit 2 that predicts the occurrence probability of a plurality of different results that may occur in the future based on past information, and calculates the occurrence probability of a plurality of different results that may occur in the future. Probabilities of occurrence of a plurality of different results that may occur in the future are sent to the plurality of terminals 91 by wired communication or wireless communication. The terminal 91 has a priority calculation unit 92 that calculates the priority of a plurality of control targets based on the occurrence probability of a plurality of different results, information detected by the terminal 91, and parameters calculated by the terminal 91. And calculate the priority of the control target. Then, based on the control target with the highest priority, the control signal is calculated by the control signal calculation means 93 that calculates the control signal for determining the operation of the terminal 91. Further, a terminal parameter update unit 94 that updates the parameters for each terminal 91 is provided, and the parameters of the priority calculation unit 92 are updated.

FIG. 2 is a system diagram of the server 1, which is the same as that of the first embodiment and will not be described in detail.
FIG. 3 is a system diagram of the terminal 3, which is the same as in the first embodiment, and will not be described in detail.

The details of each process are described below.

<Occurrence probability predicting means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information (FIG. 4)>
In this process, the probability of occurrence of a plurality of different results that may occur in the future is calculated based on past information. Specifically, although shown in FIG. 4, it is the same as that of the first embodiment, and will not be described in detail.

FIG. 22 shows an overall view of processing at the terminal 95.
The priority calculation means 96 for calculating the priority of the plurality of control targets calculates the priority of the movement route that is the priority of the control target.
The control signal calculation means 97 that calculates the control signal calculates a control signal for traveling on the determined moving route, and controls the moving body 32 that is a control target.
The terminal parameter update unit 98 that updates the parameters for each terminal 91 updates the parameters of the priority calculation unit 96.

Next, details of the priority calculation means 96, the control signal calculation means 97, and the terminal parameter update means 98 will be described.

<Priority calculating means for calculating priorities of a plurality of control targets (FIG. 23)>
In this process, priorities of a plurality of control targets are calculated. Specifically, it is shown in FIG.
The four probabilities calculated by the occurrence probability predicting means 31 (see FIG. 4) and the information detected by the terminal 91 are, for example, the following values that can be detected by a sensor or an input device installed in the terminal 91. .
・ Used years (total mileage)
・ Today's total mileage ・ Driver characteristics (gender / age / driving history / fatigue)
・ The location of the vehicle and the remaining fuel.

In addition, since the detection method of each said information is described in many literatures and books, it does not elaborate here.

The priority calculation means 99 for calculating the priority for selecting the route A multiplies the above four probabilities and the information detected by the terminal 91 by weighting factors a_11, a_12,. The priority for selecting is calculated.

In the priority calculation means 100 for calculating the priority for selecting the route B, the above four probabilities and the information detected by the terminal 91 are multiplied by weighting factors a_21, a_22,. The priority for selecting is calculated.

Further, the weighting coefficients a_11, a_12,..., A_1n and a_21, a_22,..., A_2n are sequentially determined by the update values calculated by the terminal parameter updating means 98 for updating the parameters described later for each terminal 91, Updated.

<Terminal Parameter Updating Unit 98 (FIG. 24) for Updating Parameters for Each Terminal>
In this process, the update values of the weight coefficients a_11, a_12,..., A_1n and a_21, a_22,. Specifically, it is shown in FIG.

Weight coefficients a_11, a_12,... Are minimized so that the error between “value of A1 below” and “value obtained by performing the same calculation as the priority calculation means 99 for selecting path A using B below” is minimized.・ Calculate the update value of a_1n.
A1:
-Frequency A2 for selecting route A when operating manually:
-Frequency B for selecting route B when operating manually:
-Probability of arriving at the target value at the time when route A is selected-Probability of arriving at the target value at the time when route B is selected-When the route A is selected, the fuel consumption deterioration rate is 10 Probability that the fuel efficiency deterioration rate will be 10% or more when route B is selected. • Years of use (total mileage)
・ Today's total mileage ・ Driver characteristics (gender / age / driving history / fatigue)
・ Vehicle position ・ Fuel remaining amount.

Also, the weight coefficients a_21, a_22, and so on are minimized so that the error between “the value of A2” and “the value obtained by performing the same calculation as the priority calculation means 100 for selecting the path B using B” is minimized. ..., calculate the updated value of a_2n.

As a method of determining the weighting factor, linear regression using the least square method or logistic regression can be considered. Details of these methods have been described in many documents and books, and will not be described in detail here.
In addition to linear regression and logistic regression, other learning methods such as reinforcement learning may be used.

Although the occurrence probability prediction means (see FIG. 4) for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information is a Bayesian network, a similar function can be realized as a neural network.
Further, in the priority calculation means (see FIG. 23) for calculating the priority of a plurality of control targets, the priority of the route to be selected is calculated, but a configuration in which the priority of the fuel consumption deterioration rate is also considered.

Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the above-described embodiments, and various designs can be made without departing from the spirit of the present invention described in the claims. It can be changed. For example, the above-described embodiment has been described in detail for easy understanding of the present invention, and is not necessarily limited to one having all the configurations described. Further, a part of the configuration of an embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of an embodiment. Furthermore, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

DESCRIPTION OF SYMBOLS 1 Server 2 Occurrence probability prediction means 3 Terminal 4 Priority calculation means 5 Control signal calculation means 11 Server storage device 12 Server CPU
13 Server ROM
14 Server RAM
15 Server Data Bus 16 Server Input Circuit 17 Server Input / Output Port 18 Server Output Circuit 21 Terminal Storage Device 22 Terminal CPU
23 Terminal ROM
24 Terminal RAM
25 Terminal Data Bus 26 Terminal Input Circuit 27 Terminal Input / Output Port 28 Terminal Output Circuit 31 Occurrence Probability Prediction Unit 32 Mobile Object (Autonomous Car, etc.)
33 Terminal 34 Priority calculation means 35 Control signal calculation means 36 Priority calculation means 37 for selecting the route A Priority calculation means 38 for selecting the route B Calculation means 39 for determining the route with the highest priority as the movement route 39 The determined movement route Calculation means 41 for calculating a control signal for traveling on the vehicle Generation probability prediction means 42 Engine (prime mover)
43 Fuel injection valve 44 Electronic throttle 45 Spark plug 46 Terminal 47 Priority calculation means 48 Control signal calculation means 49 Priority calculation means 50 for selecting a power mode Priority calculation means 51 for selecting a fuel consumption mode Priority calculation means 52 for selecting an intermediate mode Calculation means 53 for determining the highest priority mode as the operation mode Calculation means 61 for calculating the control signal for setting the determined operation mode Occurrence probability prediction means 62 Robot 63 Terminal 64 Priority calculation means 65 Control signal calculation means 66 Priority calculation means 67 for selecting a speed mode Priority calculation means 68 for selecting a careful mode Priority calculation means 69 for selecting an intermediate mode Calculation means 70 for determining the mode with the highest priority as the operation mode 70 for making the determined operation mode Calculation means 71 for calculating the control signal Priority display for selecting path A Calculating means 72 Priority calculating means 81 for selecting path B Server 82 Occurrence probability predicting means 83 Parameter value updating means 84 Terminal 85 Priority calculating means 86 Control signal calculating means 87 Occurrence probability predicting means 88 Bayesian network parameter value updating means (Bayes Updater)
91 Terminal 92 Priority calculation means 93 Control signal calculation means 94 Terminal parameter update means 95 Terminal 96 Priority calculation means 97 Control signal calculation means 98 Parameter update means 99 Priority calculation means for selecting path A 100 Priority for selecting path B , A_1n identification processing 102 Parameters a_21, a_22,..., A_2n identification processing

Claims (17)

A control device comprising a server and a plurality of terminals,
An occurrence probability prediction means for predicting the occurrence probability of a plurality of different results that may occur in the future based on past information;
at least,
The probability of occurrence of the plurality of different results;
Information detected at the terminal;
Based on the parameters calculated at the terminal,
Priority calculation means for calculating the priority of a plurality of control targets, and at least
Based on the control priority with the highest priority,
A control apparatus comprising: control signal calculation means for calculating a control signal for determining the operation of the terminal. Parameters calculated in the terminal are:
at least,
In the occurrence probability of the plurality of different results and the information detected in the terminal,
A parameter for weighting,
The control device according to claim 1, wherein the priority calculation unit calculates the priority of a plurality of control targets based on the weighted value. The occurrence probability prediction means includes:
The control device according to claim 1, further comprising parameter value updating means for updating a parameter value of the occurrence probability prediction means based on newly obtained information. The control device according to claim 1, further comprising terminal parameter update means for updating a parameter calculated in the terminal for each terminal. Information detected at the terminal is:
The control device according to claim 1, wherein the control device is a value based on at least a sensor output value of a sensor installed in the terminal. Information detected at the terminal is:
The control device according to claim 1, wherein the control device is a value that is uniquely learned based on at least a sensor output value of a sensor installed in the terminal. The occurrence probability prediction means includes:
The control device according to claim 1, wherein the control device is at least a Bayesian network or a neural network. The priority calculation means includes:
The control device according to claim 1, wherein the control device is at least logistic regression or reinforcement learning. The occurrence probability prediction means includes:
Based on newly obtained information,
At least by Bayesian update or stochastic gradient descent
2. The control apparatus according to claim 1, further comprising parameter value updating means for updating a parameter value of the occurrence probability prediction means. The control target of the control device is
It is a mobile object such as an autonomous driving car
The control device according to claim 1, wherein the plurality of control targets are movement paths of the moving body. The control target of the control device is
An internal combustion engine,
The control device according to claim 1, wherein the plurality of control targets are operating conditions of the internal combustion engine. The control target of the control device is
A robot,
The control device according to claim 1, wherein the plurality of control targets are tasks performed by the robot. The occurrence probability prediction means includes:
Based on information newly obtained from the terminal,
The control apparatus according to claim 1, further comprising: a parameter value updating unit that updates a parameter value of the occurrence probability prediction unit. The control apparatus according to claim 1, wherein a plurality of different results that may occur in the future are set as the plurality of control targets. 2. The control apparatus according to claim 1, wherein the current information is used as input information of the occurrence probability prediction means. The probability of multiple different outcomes that may occur in the future is at least the probability that the mobile will arrive at the target value on time, or
Probability that the energy efficiency / fuel efficiency of the moving body falls within the specified range, or
Probability that the motor / engine torque can exceed the specified value, or
The probability that the fuel efficiency improvement rate of the prime mover / engine will be above the specified value, or
The probability that the fuel efficiency deterioration rate of the prime mover / engine will be greater than or equal to a predetermined value, or
The control apparatus according to claim 1, wherein a probability that a robot task is completed within a predetermined time or a probability that a robot risk level is increased is defined. A control device comprising a server and a plurality of terminals,
The server is provided with an occurrence probability prediction means for predicting an occurrence probability of a plurality of different results that may occur in the future based on past information.

Images