Awakening degree control device, awakening degree control method, and program

Abstract

The alertness control device comprises: a physical quantity prediction model which is an explicit function that includes physical quantities of the surrounding environment that affect the alertness of the subject and setting values ​​of a control device that affects the physical quantities as explanatory variables, and has the predicted values ​​of the physical quantities as explained variables; an alertness prediction model which is an explicit function that includes the physical quantities and their changes over time as explanatory variables, and has the predicted values ​​of the changes in the alertness over time as explained variables; constraint conditions that include a setting value range condition that the setting value is within a predetermined range; and an objective function that represents the sum or average of predicted values ​​of one or more subjects including the subject, and the one or more subjects whose predicted values ​​of the changes in the alertness for two or more time steps satisfy the predetermined condition; a setting value calculation means that calculates the setting value so that the value of the objective function is maximized under the constraint conditions; and a setting means that sets the calculated setting value in the control device.

Description

Technical Field

[0001] The present invention relates to an arousal level control device, an arousal level control method, and Program the like.

Background Art

[0002] Techniques have been proposed for acquiring user's biological information and calculating the user's arousal level from the acquired biological information (for example, Patent Documents 1 and 2). Here, the arousal level is an index indicating the degree of wakefulness. The lower the value of the arousal level, the sleepier the subject is.

[0003] In a state where the arousal level is low, the work efficiency often decreases when the user performs work. Therefore, a state with a low arousal level is not an appropriate state for work performance. For example, in office work, the work efficiency decreases, and in automobile driving, distracted driving increases. Thus, a state with a low arousal level tends to be an undesirable state for various tasks.

[0004] Therefore, systems have been proposed for improving the arousal level or controlling the user's environment so as to be within an appropriate range (Patent Documents 3, 4, and 5).

[0005] Patent Document 3 discloses a system for controlling the arousal level for vehicle drivers, which changes the settings of environmental control devices such as air conditioners and lighting to predetermined settings when the predicted value of the user's arousal level when the current environmental state continues is lower than a predetermined threshold.

[0006] Patent Document 4 discloses a system for controlling the arousal level for vehicle drivers, which determines and controls the combination and intensity of devices that stimulate the five senses such as air conditioners and lighting based on predetermined settings according to the position of the user's current state in a two-dimensional coordinate system composed of a sleepiness-arousal evaluation axis and a pleasure-displeasure evaluation axis, particularly according to how far the user's current state is from a desired range.

[0007] Patent Document 5 discloses a system for controlling the arousal level for vehicle drivers, which gives the user a thermal and cold stimulus due to temperature changes by periodically switching an air conditioning device to a predetermined operation mode (temperature, air volume setting) when the arousal level of the target person falls below a preset threshold value.

[0008] There is also a technology for acquiring and processing user information or information on the surrounding environment. For example, the mood estimation system of Patent Document 6 indexes the mood based only on the heart rate of the target person, and when the index value deviates from a preset range, indexes the mood of the target person based on a plurality of biological information of the target person and a plurality of environmental information of the surrounding environment of the target person.

[0009] Also, the air conditioning management system described in Patent Document 7 calculates a predicted environmental value after a predetermined time based on the environmental value detected by a detection device, calculates parameters of the air conditioning device based on the environmental value and the predicted environmental value, and transmits them to the air conditioning device.

[0010] Also, in the arousal level maintenance method described in Patent Document 8, the arousal level is detected from deep body temperature such as the tympanic membrane temperature of the operator, and when a decrease in the arousal level of the operator is observed, the illuminance is changed from an illuminance suitable for work to a higher illuminance to give the operator an arousal effect by light stimulation.

[0011] Also, the drowsiness estimation device described in Patent Document 9 includes a neural network having a two-layer structure of an image processing neural network and a drowsiness estimation neural network. The image processing neural network estimates the age and gender of the user, and extracts specific actions and states of the user indicating a drowsy state such as closing the eyes. The drowsiness estimation neural network determines the drowsy state of the user based on the extraction results of specific actions and states of the user indicating a drowsy state, the detection results of an indoor environment information sensor, and taking into account the age and gender of the user.

[0012] This Patent Document 9 describes that the control unit of an air conditioner calculates the air conditioning control content so that the estimated sleepiness level is equal to or lower than a threshold value, and executes the calculated air conditioning control. Furthermore, Patent Document 9 describes that when no desired change is observed in the actions and states of the user, the estimated sleepiness state may deviate from the actual sleepiness state, so the estimation model is updated.

Prior Art Documents

Patent Documents

[0013]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Patent Document 5

Patent Document 6

Patent Document 7

Patent Document 8

Patent Document 9

Summary of the Invention

Problems to be Solved by the Invention

[0014] When an apparatus or system exerts an influence on the surrounding environment of a person to be controlled for wakefulness and performs wakefulness control, in order to perform wakefulness control with high precision, it is preferable that the influence of the action on the surrounding environment on wakefulness can be grasped more accurately.

[0015] An example of an object of the present invention is an arousal control device, an arousal control method, and Program to provide

Means for Solving the Problems

[0016] According to a first aspect of the present invention, an arousal control device a set value set as a control target value of a physical quantity for at least any one of the temperature, illuminance, sound volume, and vibration magnitude of the surrounding environment of the person to be subjected to arousal control, or a combination thereof, or a physical quantity A physical quantity prediction model that is a model including, as explanatory variables, a set value that affects a control device, and is a positive function in which, in a state where the set value is set in the control device, the physical quantity is a predicted value of the physical quantity after a predetermined time has elapsed from the timing when the physical quantity becomes a measurement value measured by an environmental measurement device for measuring the physical quantity, and the model parameters are identified by machine learning. within the range of the constraint condition for the set value, that is, included in a predetermined range set by the user the physical quantity and the set value A wakefulness prediction model that includes, as explanatory variables, the physical quantity and its time change amount and an estimated value of the wakefulness, and is a positive function in which a predicted value of the time change amount of the wakefulness is a response variable, and the model parameters are identified by machine learning. rule a set value calculation means for performing a solution search for the set value so that the value of an objective function, which is a function of the sum value or average value of the predicted values of the time change amount of the wakefulness of the one or more subjects over one or more intervals of time steps, becomes larger, including as part of a function a setting means for setting the calculated set value in the control device, calculate and objective function is provided.

[0017] According to a second aspect of the present invention, an arousal control method by a computer, a physical quantity that is at least any one of the temperature, illuminance, sound volume, and vibration magnitude of the surrounding environment of the person to be subjected to arousal control, or a combination thereof, or a physical​​quantity A set value set as a control target value of the physical quantity for a control device that has an impact within the range of the constraint condition for the set value, that is, included in a predetermined range set by the user the physical quantity and the set value A physical quantity prediction model that is a model including the set value as an explanatory variable, and is a positive function in which the predicted value of the physical quantity after a predetermined time has elapsed from the timing when the physical quantity has become the measured value by the environmental measurement device that measures the physical quantity in the state where the set value is set in the control device is the explained variable, and is a physical quantity prediction model obtained by identifying model parameters by machine learning rule A model that includes the physical quantity and its time change amount and an estimated value of the arousal level as explanatory variables, and is a positive function in which the predicted value of the time change amount of the arousal level is the explained variable, and is an arousal level prediction model obtained by identifying model parameters by machine learning including as part of a function The sum value or average value of the predicted values of the time change amount of the arousal level of one or more of the subjects, and for one or more time steps calculate Of the objective function that is a function objective function Perform a solution search for the set value so that the value becomes larger, Including setting the calculated set value in the control device

[0018] According to a third aspect of the present invention, the program causes A computer to At least any one of the temperature, illuminance, sound volume, and vibration magnitude of the surrounding environment of a subject of arousal control, or a physical combination of these quantity A set value set as a control target value of the physical quantity for a control device that has an impact within the range of the constraint condition for the set value, that is, included in a predetermined range set by the user the physical quantity and the set value A physical quantity prediction model that is a model including the set value as an explanatory variable, and is a positive function in which the predicted value of the physical quantity after a predetermined time has elapsed from the timing when the physical quantity has become the measured value by the environmental measurement device that measures the physical quantity in the state where the set value is set in the control device is the explained variable, and is a physical quantity prediction model obtained by identifying model parameters by machine learning ruleA wakefulness prediction model represented by a positive function that includes the physical quantity, its time change amount, and the estimated value of the wakefulness as explanatory variables, and uses the predicted value of the time change amount of the wakefulness as the response variable, and the model parameters are identified by machine learning. including as part of a function The sum or average value of the predicted values of the time change amount of the wakefulness of the subject for one or more subjects and for one or more intervals of time steps. calculate Of the objective function that is a function. objective function Search for a solution of the set value so that the value becomes larger. Set the calculated set value to the control device. A program for causing this to be executed.

Advantages of the Invention

[0019] According to the embodiment of the present invention, when controlling wakefulness, it is possible to more accurately grasp the influence of the action on the surrounding environment on wakefulness.

Brief Description of the Drawings

[0020]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Mode for Carrying Out the Invention

[0021] Hereinafter, embodiments of the present invention will be described. However, the following embodiments do not limit the invention according to the claims. Also, not all combinations of features described in the embodiments are essential for the solution means of the invention. FIG. 1 is a schematic block diagram showing an example of the device configuration of an arousal level control system 1 according to an embodiment. In the configuration shown in FIG. 1, the arousal level control system 1 includes an arousal level control device 100, one or more environmental control devices 200, one or more environmental measurement devices 300, and one or more arousal level estimation devices 400.

[0022] The arousal level control device 100 is connected to each of the environmental control devices 200, each of the environmental measurement devices 300, and each of the arousal level estimation devices 400 via the communication line 900, and is capable of communicating with these devices. The communication line 900 may be configured in any form regardless of the form of occupation of the communication line such as a dedicated line, the Internet, a VPN (Virtual Private Network), a LAN (Local Area Network), etc., and the physical form of the communication line such as a wired line or a wireless line.

[0023] The arousal level control system 1 determines the arousal level of the person subject to arousal level control, and controls the physical quantity of the surrounding environment of the person subject to arousal level control according to the determination result, in order to maintain or improve the arousal level. As described above, the arousal level is an index indicating the degree of being awake. The lower the value of the arousal level, the sleepier the person subject to arousal level control is. The person subject to arousal level control is also referred to as the user or simply the subject.

[0024] Here, the physical quantity of the surrounding environment of the subject is a physical quantity (physical amount) that affects the subject, and here in particular, it is a physical quantity that affects the arousal level of the subject. The physical quantity of the surrounding environment of the subject is also simply referred to as the physical quantity. Examples of physical quantities include air temperature such as room temperature, and brightness such as illuminance by lighting equipment, but are not limited to these. For example, the arousal level control system 1 may give stimuli other than temperature and brightness, such as moisture (humidity), sound, or vibration, to the subject in addition to or instead of temperature and brightness. Also, the arousal level control system 1 may use magnitudes such as moisture (humidity), sound, or vibration as physical quantities.

[0025] Hereinafter, the air temperature is simply referred to as temperature. However, the arousal level control system 1 may control other temperatures in addition to or instead of the air temperature. For example, the arousal level control system 1 may control the temperature of an object that directly contacts the subject. As a specific example, a heater may be provided on the seat surface of the subject's seat, and the arousal level control system 1 may control the temperature of the heater.

[0026] The unit by which the arousal control system 1 controls the physical quantity is not limited to a specific one. For example, a spot-type air conditioner (local air conditioner) and a lighting stand may be installed on an individual's seat, and the arousal control system 1 may control the physical quantity in units of seats. Alternatively, the arousal control system 1 may control the physical quantity in units of rooms or may control the physical quantity of the entire building. Further, when controlling the physical quantity of the entire building, the target person does not have to be all the people in the building, and may be some of the people in the building.

[0027] The number of target persons may be one or a plurality. Only specific persons such as when the arousal control system 1 accepts registration of the target person may be the target persons. Alternatively, unspecified persons located in the control target space of the arousal control system 1 may be the target persons. When there are a plurality of target persons, the arousal control system 1 may control the physical quantity for each target person or may control the physical quantity in common for a plurality of target persons.

[0028] In order to improve the arousal level of the target person, it is conceivable to control the physical quantity so that the comfort level decreases for some people, such as raising the room temperature or brightening the lighting. The arousal control system 1 can balance ensuring the arousal level of the target person and comfort by determining the arousal level of the target person for arousal control and controlling the physical quantity according to the determination result. For example, the arousal control system 1 may control the physical quantity so as to improve the arousal level only when the arousal level of the target person has decreased.

[0029] Hereinafter, the case where the arousal control system 1 improves the arousal level of the subject (awakens drowsiness) will be described as an example, but it is not limited to such an example. For example, the arousal control system 1 may lower the arousal level of the subject (induce sleep). For example, the arousal control system 1 may switch between control for improving the arousal level and control for lowering the arousal level according to the time zone and execute them. Alternatively, when it is predicted that the arousal level of the subject will decrease, the arousal control system 1 may control so that the arousal level of the subject does not decrease (that is, so that the subject does not doze off). Alternatively, when it is predicted that the arousal level of the subject will increase, the arousal control system 1 may control so that the arousal level of the subject does not increase (that is, so that the subject does not awaken).

[0030] The arousal control device 100 controls the environmental control device 200 according to the arousal level of the subject. By controlling the environmental control device 200, the arousal control device 100 controls the physical quantity of the surrounding environment of the subject, thereby controlling the arousal level of the subject. The arousal control device 100 is configured using a computer such as a personal computer (PC) or a workstation.

[0031] The environmental control device 200 is a device that adjusts physical quantities. As described above, examples of physical quantities include, for example, air temperature and illuminance. The temperature can be adjusted by an air conditioner, and the illuminance can be adjusted by a lighting device. Thus, examples of the environmental control device 200 include, but are not limited to, an air conditioner and a lighting device. The environmental control device 200 corresponds to an example of a controlled device and is controlled by the arousal control device 100 as described above.

[0032] Devices other than the environmental control device 200, such as the arousal control device 100, can acquire information regarding the operating state, such as device setting values, from the environmental control device 200, and can update the device setting values for the environmental control device 200. Here, the device setting value is a physical quantity set in the environmental control device 200 as a control target value. The device setting value is also referred to as a set value of a physical quantity, or simply a set value. When the environmental control device 200 is an air conditioner, the set temperature can be used as the device setting value. When the environmental control device 200 is a lighting device, lighting output (for example, luminous intensity, illuminance, current value, power value, etc.) can be used as the device setting value. In the following, the case where illuminance is used as the device setting value of the lighting device will be described as an example, but it is not limited thereto.

[0033] The environmental measurement device 300 is a device that measures physical quantities such as temperature and illuminance and converts them into numerical data. Examples of the environmental measurement device 300 include, but are not limited to, a temperature sensor and an illuminance sensor.

[0034] The arousal estimation device 400 is a device that estimates the arousal level of a subject from biological information or the like and converts it into numerical data. The arousal estimation device 400 may use, but is not limited to, any one of body temperature, facial video, and pulse wave, or a combination thereof, as biological information. The arousal estimation device 400 measures or calculates biological information and converts the obtained biological information into a numerical value (arousal level) indicating the degree of arousal. The arousal estimation device 400 is not essential for the arousal control system 1. When the arousal control system 1 does not include the arousal estimation device 400, the arousal of the subject is estimated based on the physical quantity.

[0035] Next, the functional configuration of the arousal control device 100 will be described. FIG. 2 is a schematic block diagram showing an example of the functional configuration of the arousal level control device 100. In the configuration shown in FIG. 2, the arousal level control device 100 includes a communication unit 110, a storage unit 170, and a control unit 180. The storage unit 170 includes (stores) a physical quantity prediction model 171 and an arousal level prediction model 172. The control unit 180 includes a monitoring control unit 181, a first acquisition unit 182, a second acquisition unit 183, a set value calculation unit 184, a physical quantity prediction model calculation unit 185, an arousal level prediction model calculation unit 186, a set value determination unit 187, a physical quantity prediction model learning unit 188, and an arousal level prediction model learning unit 189.

[0036] The communication unit 110 communicates with other devices according to the control of the control unit 180. In particular, the communication unit 110 receives various information from each of the environment control device 200, the environment measurement device 300, and the arousal level estimation device 400. Further, the communication unit 110 transmits device set values to the environment control device 200. The storage unit 170 stores various information. The storage unit 170 is configured using a storage device included in the arousal level control device 100.

[0037] The physical quantity prediction model 171 is a mathematical model that calculates a predicted value of the physical quantity based on a set value (device set value) of the physical quantity. More specifically, the physical quantity prediction model 171 calculates a predicted value of the physical quantity when a predetermined time has elapsed based on the measured value of the physical quantity measured by the environment measurement device 300 and the set value of the physical quantity set in the environment control device 200.

[0038] When the above-mentioned predetermined time has elapsed refers to the time after a predetermined time has elapsed from the time when the physical quantity given to the physical quantity prediction model 171 is measured. Instead of the measurement time of the physical quantity given to the physical quantity prediction model 171, the time when the arousal level control device 100 (communication unit 110) receives the physical quantity can be used. The above-mentioned predetermined time may be fixed at a certain time or may be variable as a model parameter. The model parameter here is a set parameter of the physical quantity prediction model 171. The value of the model parameter is referred to as the model parameter value.

[0039] The arousal prediction model 172 is a mathematical model that calculates a predicted value of arousal based on the predicted value of the physical quantity calculated by the physical quantity prediction model 171. Furthermore, the arousal prediction model 172 calculates a predicted value of arousal based on the amount of change in the physical quantity in addition to the predicted value of the physical quantity. More specifically, the arousal prediction model 172 uses the history of the predicted values of the physical quantity calculated by the physical quantity prediction model 171, and based on the time average value and the amount of change of the physical quantity, calculates a predicted value of the amount of change in the arousal of the subject when a predetermined time has elapsed. The arousal prediction model 172 may calculate a predicted value of arousal based at least on the temporal variation of arousal.

[0040] The control unit 180 controls each part of the arousal control device 100 to execute various processes. The control unit 180 is realized by a CPU (Central Processing Unit, central processing unit) provided in the arousal control device 100 reading a program from the storage unit 170 and executing it. The monitoring control unit 181 communicates with the environment control device 200 via the communication unit 110. In the communication with the environment control device 200, the monitoring control unit 181 acquires the device setting values set in the environment control device 200. Also, the monitoring control unit 181 updates the device setting values of the environment control device 200 in the communication with the environment control device 200. For example, the monitoring control unit 181 communicates with the environment control device 200 at regular intervals, and saves the device setting values acquired in the communication together with the time stamp at the time of acquisition (reception). The storage here means, for example, storing in the storage unit 170. In this way, the monitoring control unit 181 sets device setting values for the controlled device. The monitoring control unit 181 corresponds to an example of a setting unit (setting means).

[0041] The monitoring and control unit 181 sets, as the device set value, the device set value calculated by the set value calculation unit 184 or the device set value determined by the set value determination unit 187 in the environmental control device 200. When the set value calculation unit 184 can calculate the device set value and the calculated device set value satisfies a predetermined condition (the calculated device set value is equal to or greater than the predetermined condition), that is, when it is determined that the calculated device set value is of high precision, the monitoring and control unit 181 sets the device set value calculated by the set value calculation unit 184 in the environmental control device 200. On the other hand, when the set value calculation unit 184 cannot calculate the device set value, or when the device set value calculated by the wakefulness control system 1 does not satisfy a predetermined condition, that is, when it is determined that the calculated device set value is of low precision, the monitoring and control unit 181 sets the device set value determined by the set value determination unit 187 in the environmental control device 200.

[0042] For example, when the set parameter value of the physical quantity prediction model 171 is not set, the physical quantity prediction model 171 cannot calculate the predicted value of the physical quantity, and thus it is conceivable that the set value calculation unit 184 cannot calculate the device set value. Also, when the set parameter value of the wakefulness prediction model 172 is not set, the wakefulness prediction model 172 cannot calculate the predicted value of the wakefulness, and thus it is conceivable that the set value calculation unit 184 cannot calculate the device set value.

[0043] Also, when the prediction accuracy of the physical quantity by the physical quantity prediction model 171 drops below a predetermined condition, it is conceivable that the accuracy of the device set value calculated by the set value calculation unit 184 decreases. Also, when a predetermined time or more has elapsed since the setting of the set parameter value of the physical quantity prediction model 171, the prediction accuracy of the physical quantity by the physical quantity prediction model 171 decreases, and it is conceivable that the accuracy of the device set value calculated by the set value calculation unit 184 also decreases.

[0044] Also, when the prediction accuracy of the arousal level by the arousal level prediction model 172 drops below a predetermined condition, it is conceivable that the accuracy of the device setting value calculated by the setting value calculation unit 184 decreases accordingly. Also, when a predetermined time or more has elapsed since the setting of the setting parameter value of the arousal level prediction model 172, the prediction accuracy of the arousal level by the arousal level prediction model 172 decreases, and it is conceivable that the accuracy of the device setting value calculated by the setting value calculation unit 184 also decreases. In some or all of these cases, the monitoring control unit 181 may set the device setting value determined by the setting value determination unit 187 in the environmental control device 200.

[0045] The first acquisition unit 182 communicates with the environmental measurement device 300 via the communication unit 110 and acquires the measurement value of the physical quantity measured by the environmental measurement device 300. For example, the first acquisition unit 182 communicates with the environmental measurement device 300 at regular intervals and stores the measurement value of the physical quantity acquired by the communication together with the time stamp at the time of acquisition (reception). This time stamp can be regarded as indicating the time when the environmental measurement device 300 measures the physical quantity.

[0046] The second acquisition unit 183 communicates with the arousal level estimation device 400 and acquires the estimated value of the arousal level of the subject. For example, the second acquisition unit 183 communicates with the arousal level estimation device 400 at regular intervals and stores the estimated value of the arousal level acquired by the communication together with the time stamp at the time of acquisition (reception). This time stamp can be regarded as indicating the estimation time of the arousal level by the arousal level estimation device 400. The estimated value of the arousal level of the subject is also referred to as the arousal level estimated value.

[0047] The setting value calculation unit 184 calculates the device setting values of the environmental control device 200 so as to improve the user's wakefulness. For example, the setting value calculation unit 184 calculates the device setting values at regular intervals. The setting value calculation unit 184 acquires the device setting values from the monitoring control unit 181, acquires the measured values of the physical quantities from the first acquisition unit 182, and acquires the wakefulness estimation values from the second acquisition unit 183, and calculates the device setting values based on these. The setting value calculation unit 184 outputs the calculated device setting values to the monitoring control unit 181. The monitoring control unit 181 sets the device setting values for the environmental control device 200 by transmitting the device setting values acquired from the setting value calculation unit 184 to the environmental control device 200 via the communication unit 110.

[0048] The setting value calculation unit 184 uses the physical quantity prediction model 171 and the wakefulness prediction model 172 to solve (or approximately solve) an optimization problem under the constraint conditions regarding the physical quantities, thereby calculating the setting values for controlling the wakefulness of the subject. The setting value calculation unit 184 calculates the device setting values so that the wakefulness becomes higher by solving (or approximately solving) the optimization problem. In this way, the process of the setting value calculation unit 184 solving the optimization problem corresponds to an example of a process of making the objective function value such as wakefulness higher (or lower, or closer to the target value). The setting value calculation unit 184 may calculate the device setting values when the wakefulness is the highest by solving (or approximately solving) the optimization problem.

[0049] In the optimization problem solved by the setting value calculation unit 184, the physical quantity prediction model 171 is used as the first constraint condition, the wakefulness prediction model 172 is used as the second constraint condition, and the condition that the device setting values of the environmental control device 200 are within a predetermined range is used as the third constraint condition. The setting value calculation unit 184 solves the optimization problem including these constraint conditions. The predetermined range of the device setting values here is the settable range defined by the specifications of the environmental control device 200.

[0050] In addition, the objective function of the optimization problem solved by the setting value calculation unit 184 is, for example, a function that calculates the total value or average value of the predicted values of the change amounts of the arousal levels for one or more (or two or more) subjects and for one or more intervals of time steps. The setting value calculation unit 184 solves the optimization problem so as to increase the value of this objective function and calculates the device setting values. The setting value calculation unit 184 may calculate the device setting values when this objective function is at its maximum. The total value of the predicted values of the change amounts of the arousal levels may be the sum of the predicted values of the change amounts of the arousal levels for each subject. The average value of the predicted values of the change amounts of the arousal levels may be a value obtained by dividing the sum of the predicted values of the change amounts of the arousal levels for each subject by the number of subjects. The optimization problem solved by the setting value calculation unit 184 is referred to as an arousal optimization model. The arousal optimization problem is configured as a mathematical model.

[0051] The setting value calculation unit 184 may calculate the setting values so that the trimmed average value of the predicted values of the change amounts of the arousal levels for one or more subjects and for one or more intervals of time steps becomes larger. By using the trimmed average, the setting value calculation unit 184 can optimize the whole because, for example, when there are subjects among the subjects whose change in arousal level is extremely small or, conversely, extremely large with respect to the change in the physical quantity, those extreme subjects are not over-evaluated. Or, this optimization may be performed for some of the subjects.

[0052] The setting value calculation unit 184 may solve an optimization problem including a constraint condition regarding the comfort score calculated for the device setting values and calculate the device setting values that satisfy this constraint condition. For example, the setting value calculation unit 184 may calculate each of a plurality of types of device setting values so that the total value of the comfort penalty scores calculated for each of the plurality of types of device setting values is within a predetermined range. In other words, the setting value calculation unit 184 may calculate each of the plurality of types of device setting values when the condition that the comfort score is included in a certain range is satisfied. In this way, by calculating the device setting value so that the setting value calculation unit 184 satisfies the constraint conditions related to comfort, it is possible to prevent the comfort from extremely decreasing.

[0053] The physical quantity prediction model calculation unit 185 reads and executes the physical quantity prediction model 171 from the storage unit 170. Therefore, the physical quantity prediction model calculation unit 185 executes the prediction of the physical quantity using the physical quantity prediction model 171. The arousal level prediction model calculation unit 186 reads and executes the arousal level prediction model 172 from the storage unit 170. Therefore, the arousal level prediction model calculation unit 186 executes the prediction of the arousal level using the arousal level prediction model 172.

[0054] When the setting value calculation unit 184 cannot calculate a device setting value that can obtain a high arousal effect, the setting value determination unit 187 calculates the device setting value instead of the setting value calculation unit 184 and outputs it to the monitoring control unit 181. One of the purposes is to generate learning data that allows the physical quantity prediction model learning unit 188 and the arousal level prediction model learning unit 189 to perform learning efficiently, as will be described later.

[0055] The physical quantity prediction model learning unit 188 sets or updates the physical quantity prediction model 171 by obtaining the setting parameter values of the physical quantity prediction model 171 through machine learning or the like. The physical quantity prediction model learning unit 188 performs machine learning or the like when at least one of the following cases occurs: when the setting parameter values of the physical quantity prediction model 171 are not set, when the prediction accuracy by the physical quantity prediction model 171 drops below a predetermined condition, and when more than a predetermined time has elapsed since the setting of the setting parameter values of the physical quantity prediction model 171.

[0056] The physical quantity prediction model learning unit 188 acquires the device setting value from the monitoring and control unit 181 and the measured value of the physical quantity from the first acquisition unit 182 as learning data. The physical quantity prediction model learning unit 188 performs machine learning or the like based on these measured values of the physical quantity and the set value (device setting value) of the physical quantity to obtain the set parameter values of the physical quantity prediction model 171. The physical quantity prediction model learning unit 188 outputs the parameter values of the physical quantity prediction model 171 obtained by machine learning or the like to the set value calculation unit 184 and the set value determination unit 187.

[0057] The arousal prediction model learning unit 189 sets or updates the arousal prediction model 172 by acquiring the set parameter values of the arousal prediction model 172 through machine learning or the like. When the set parameter values of the arousal prediction model 172 are not set, when the prediction accuracy by the arousal prediction model 172 drops below a predetermined condition, and when at least one of the cases where a predetermined time or more has elapsed since the setting of the set parameter values of the arousal prediction model 172 occurs, the arousal prediction model learning unit 189 performs machine learning or the like.

[0058] For the arousal control device 100, the arousal prediction model learning unit 189 is not essential. In particular, when the arousal control system 1 does not include the arousal estimation device 400, the arousal control device 100 does not acquire the information on arousal from the outside, and thus cannot obtain the correct data in the machine learning data of arousal. In this case, it is conceivable that the arousal control device 100 is configured not to include the arousal prediction model learning unit 189. When the arousal prediction model learning unit 189 is not provided, the arousal control device 100 may continue to use the arousal prediction model 172 as it is, and when the update of the arousal prediction model 172 becomes necessary, operations such as manual update by the administrator of the arousal control system 1 may be considered. Alternatively, the arousal control device 100 may automatically update the model parameter values by a method such as acquiring the latest model parameter values via the Internet.

[0059] The arousal prediction model learning unit 189 acquires the measured values of physical quantities from the first acquisition unit 182 and the estimated arousal values from the second acquisition unit 183 as learning data. The arousal prediction model learning unit 189 performs machine learning or the like based on these measured values of physical quantities and the estimated arousal values to obtain the set parameter values of the arousal prediction model 172.

[0060] Hereinafter, examples of specific calculation procedures for each of the set value calculation unit 184, the set value determination unit 187, the physical quantity prediction model learning unit 188, and the arousal prediction model learning unit 189 will be described. First, an example of an arousal optimization model (optimization problem) used by the set value calculation unit 184 to calculate the device set value will be described. The set value calculation unit 184 calculates the device set value by performing a mathematical optimization calculation on this arousal optimization model.

[0061] In this arousal optimization model, the following constants, coefficients, variables, and functions are used. (Decision variable) T t set : Air conditioning temperature set value at time step t L t set : Lighting output set value at time step t The decision variables are the variables whose values are calculated by the set value calculation unit 184 in the optimization operation. In the case of the example described here, the set value calculation unit 184 calculates the temperature set for the environmental control device 200 which is an air conditioning device and the illuminance set for the environmental control device 200 which is a lighting device by solving the optimization problem.

[0062] (Dependent variable) A Δ : Average value of the predicted change amount of arousal with respect to the subject and time step A i Δ : Average value of the predicted change amount of arousal of subject i with respect to time step A i,t Δ : Predicted change amount of arousal of subject i at time step t T t: Predicted temperature value at time step t T t Δ : Predicted value of the time change amount of temperature at time step t

[0063] Note that the change amount in the previous interval of time step t, that is, the change amount from time step t - 1 to t, is referred to as the change amount at time step t. The time change amount is the change amount due to the passage of time (change amount over time). L t : Predicted illuminance value at time step t L t Δ : Predicted value of the time change amount of illuminance at time step t T t pnlty : Deviation degree from the comfort value of the air - conditioning temperature setting value at time step t L t pnlty : Deviation degree from the comfort value of the lighting output setting value at time step t A i,t σ : Variation degree of the time change amount of the wakefulness of subject i at time step t

[0064] (Constant · Coefficient) T: Set of indices of time steps N: Set of indices of subjects T min : Lower limit value of the air - conditioning temperature setting value T max : Upper limit value of the air - conditioning temperature setting value L min : Lower limit value of the lighting output power setting value L max : Upper limit value of the lighting output setting value

[0065] T best : Comfort value of the air - conditioning temperature setting value p T : Penalty coefficient of air - conditioning temperature L best : Comfort value of the lighting output setting value p L : Penalty coefficient of lighting output Pmax : Upper limit value of penalty score a i (τ): Estimated arousal level of subject i at relative time τ Δτ: Time step width

[0066] (Function) f A : Arousal change prediction function (arousal prediction model) f T : Temperature prediction function (one of the physical quantity prediction models) f L : Illuminance prediction function (one of the physical quantity prediction models) (Index) t: Index of time step i: Index of subject

[0067] The objective function of this arousal optimization model is shown as in Equation (1).

[0068]

Equation

[0069] A Δ (Average value of predicted arousal change over subjects and time steps) is shown as in Equation (2).

[0070]

Equation

[0071] A i Δ (Average value of predicted arousal change of subject i over time steps) is shown as in Equation (3).

[0072]

Equation

[0073] The constraint condition that the device setting value of the air conditioning device among the environmental control devices 200 is within a predetermined range is shown as in Equation (4).

[0074]

Equation

[0075] The constraint condition that the device setting value of the lighting device among the environmental control devices 200 is within a predetermined range is shown as in Equation (5).

[0076]

Equation

[0077] The constraint condition of the physical quantity prediction model 171 regarding temperature is shown as in Equation (6).

[0078]

Equation

[0079] The constraint condition of the physical quantity prediction model 171 regarding illuminance is shown as in Equation (7).

[0080]

Equation

[0081] These constraint conditions of the physical quantity prediction model 171 indicate physical constraint conditions regarding the operation of the environmental control device 200, such as the delay from when the device setting value is set in the environmental control device 200 until the physical quantity actually reaches the device setting value.

[0082] Therefore, the physical quantity prediction model 171 includes, as explanatory variables, parameters representing physical quantities of the surrounding environment that affect the subject's arousal level and parameters representing the set values of control devices that affect the physical quantities. Further, the explained variable of the physical quantity prediction model 171 is a parameter representing the predicted value of the physical quantity. Equations (6) and (7) are exemplified by a positive function in which the value of the explained variable is calculated by applying a predetermined process represented by the physical quantity prediction model 171 to the values of the explanatory variables. Note that Equations (6) and (7) do not necessarily have to be represented by a positive function.

[0083] The constraint condition of the arousal level prediction model 172 is shown as in Equation (8).

[0084]

Number

[0085] Therefore, the arousal level prediction model 172 includes, as explanatory variables, parameters representing physical quantities and parameters representing the amount of change over time of the physical quantities. Further, the explained variable of the arousal level prediction model 172 is a parameter representing the predicted value of the amount of change over time of the arousal level. Equation (8) is exemplified by a positive function in which the value of the explained variable is calculated by applying a predetermined process represented by the arousal level prediction model 172 to the values of the explanatory variables. Note that Equation (8) does not necessarily have to be represented by a positive function.

[0086] The constraint condition of the arousal level prediction model 172 indicates how the arousal level of the subject changes with respect to the physical quantity and its change. T t Δ (The predicted value of the amount of change over time of the temperature at time step t) is shown as in Equation (9).

[0087]

Number

[0088] Lt Δ (The predicted value of the amount of change in illuminance over time at time step t) is expressed as in Equation (10).

[0089]

Equation

[0090] The set value calculation unit 184 solves a mathematical programming problem of obtaining a decision variable value that maximizes an objective function representing the average value of the predicted values of the amount of change in arousal time for all users and all time steps, as shown in Equations (1) to (3), under the constraint conditions shown in Equations (2) to (10). Thereby, the set value calculation unit 184 calculates the device set value (decision variable value). The process executed by the set value calculation unit 184 can also be, for example, a process of calculating the set value so that the value of the objective function is maximized under the constraint conditions using the arousal optimization model as described above. The process executed by the set value calculation unit 184 is not necessarily limited to the process when the value of the objective function is maximized. For example, it may be a process of calculating the set value when the value of the objective function increases.

[0091] As described above, Equations (6) and (7) are constraint conditions regarding the physical quantity prediction model 171. Equations (8) to (10) are constraint conditions regarding the arousal prediction model 172. Equations (4) and (5) are constraint conditions that the device set value of the environmental control device 200 is within a predetermined range.

[0092] The options of the arousal optimization model (optimization problem) used by the set value calculation unit 184 will be described. A Δ It may be possible to use a trimmed mean value as (the average value of the predicted values of the amount of change in arousal over the subject and time steps). In this case, Equation (11) is used instead of Equation (2).

[0093]

Equation

[0094] That is, the objective function may be a combination of formulas (1), (3), and (11) instead of formulas (1) to (3). Here, trimmedmean represents the trimmed mean. The trimmed mean is the arithmetic mean obtained by discarding data from both ends of the data arranged in descending order by a determined ratio, and then calculating the mean of the remaining data. As a result, it is possible to exclude subjects with extremely changing or non-changing arousal levels from the numerical calculation of the objective function, thus preventing the calculation of device setting values that are overly tailored to a small number of specific subjects.

[0095] That is, by using the trimmed mean in the setting value calculation unit 184, for example, when there are subjects among the subjects whose arousal level changes extremely little or conversely extremely greatly with respect to the change in the physical quantity, these extreme subjects will not be overly evaluated, so that overall optimization can be achieved. The truncation ratio (the ratio of both ends combined) in the trimmed mean is preferably 10%.

[0096] In order to prevent the comfort of the subject from decreasing too much, a constraint condition regarding comfort may be included in the constraint conditions. For example, a constraint condition for the penalty score of comfort shown in formula (12) may be included.

[0097]

Number

[0098] T t pnlty (Degree of deviation from the comfort value of the air conditioning temperature setting value at time step t) is shown as in formula (13).

[0099]

Number

[0100] L t pnlty (Degree of deviation from the comfort value of the lighting output setting value at time step t) is shown as in formula (14).

[0101]

Number

[0102] Equations (12) to (14) represent a constraint condition that the sum of the comfort penalty scores is within a predetermined range (equal to or less than a predetermined magnitude). In other words, it is a constraint condition that the equipment setting values do not result in the simultaneous discomfort of air conditioning and lighting. This has the effect of avoiding a situation where the setting values of a plurality of environmental control devices, for example, an air conditioning device and a lighting device, simultaneously deviate from their most comfortable setting values respectively.

[0103] The effect of including Equations (12) to (14) in the constraint conditions will be described with reference to FIGS. 3 and 4. FIG. 3 is a diagram showing an example of equipment setting values when the constraint conditions of Equations (12) to (14) are not included in the arousal optimization model used by the setting value calculation unit 184. The horizontal axis of the graph in FIG. 3 represents the air conditioning setting value (air conditioning temperature setting value (temperature)). The vertical axis represents the lighting setting value (lighting output setting value (illuminance)).

[0104] In the example of FIG. 3, within the range that satisfies the constraint conditions of the upper limit value and the lower limit value of the temperature that can be set for the air conditioning device, and the constraint conditions of the upper limit value and the lower limit value of the illuminance that can be set for the lighting device, the equipment setting values can be arbitrarily set. Region A11 (hatched portion) indicates the range that satisfies the constraint conditions. Regarding both the air conditioning setting value and the lighting setting value, it is considered that the comfort of the subject is relatively low near the upper limit value or near the lower limit value. Therefore, when both the air conditioning setting value and the lighting setting value are set near either the upper limit value or the lower limit value, it is considered that the combined decrease in temperature comfort and illuminance comfort results in a significant decrease in comfort for the subject.

[0105] FIG. 4 is a diagram showing an example of device setting values when the wakefulness optimization model used by the setting value calculation unit 184 includes the constraint conditions of formulas (12) to (14). The horizontal axis of the graph in FIG. 4 indicates the air conditioning setting value (air conditioning temperature setting value (temperature)). The vertical axis indicates the lighting setting value (lighting output setting value (illuminance)).

[0106] In the example of FIG. 4, in addition to the constraint conditions in the case of FIG. 3, a constraint condition is provided that the total magnitude of the deviation from the comfort value of the air conditioning setting value and the deviation from the comfort value of the lighting setting value is equal to or less than a predetermined magnitude. Region A12 (hatched portion) indicates the range that satisfies the constraint condition. As a result, even within the range of the air conditioning setting value and the lighting setting value that can be set for both the air conditioning device and the lighting device, device setting values that deviate significantly from the comfort values are not set. Thereby, for the subject, it is possible to avoid a large decrease in comfort due to the combined decrease in temperature comfort and illuminance comfort. Comparing FIG. 3 and FIG. 4, the effect of adding the constraint conditions of formulas (12) to (14) is schematically shown by the change in the range region of the device setting values from region A11 (hatched portion) in FIG. 3 to region A12 (hatched portion) in FIG. 4.

[0107] When it is possible to obtain the wakefulness estimation value from the second acquisition unit 183, formula (15) may be used instead of formula (8).

[0108]

Equation

[0109] In this case, formula (16) may be included in the constraint conditions.

[0110]

Equation

[0111] A in formula (16) i,t-1 Δ can be calculated using formula (15). Also, Ai,t σ (The degree of variation in the amount of change in the wakefulness of subject i at time step t) is expressed as in Equation (17).

[0112]

Number

[0113] By using Equations (9), (10), (15) - (17) instead of Equations (8) - (10) as the wakefulness prediction model 172, the current estimated wakefulness value can be incorporated, and there is an effect of improving the prediction accuracy. Therefore, the wakefulness prediction model includes, as explanatory variables, the time average value, the amount of change over time, and the temporal variation of wakefulness. Here, std represents the standard deviation, and the temporal variation is taken as the standard deviation. Also, here it is assumed that the future temporal variation is the same as the current value (Equation 17).

[0114] A in Equation (16) i,t σ (The degree of variation in the amount of change in the wakefulness of subject i at time step t), the degree of variation in wakefulness may be included in the input to the wakefulness prediction model 172. When the degree of variation in wakefulness is large, it is considered that the subject is drowsy, and the wakefulness of the subject is considered to be relatively low. In this way, by including the degree of variation in wakefulness in the input to the wakefulness prediction model 172, it is expected that the current state of wakefulness can be grasped more accurately, and the prediction accuracy of wakefulness is expected to be improved.

[0115] The physical quantity prediction model 171, the wakefulness prediction model 172, and the wakefulness optimization model will be further described. The physical quantity prediction model 171 is a mathematical model that can calculate the predicted value of a physical quantity when a predetermined time has elapsed based on the measured value of the physical quantity and the corresponding device setting value. When the physical quantity is temperature and the corresponding environmental control device 200 is an air conditioner, the physical quantity prediction model 171 is represented by Equation (6) as described above. When the physical quantity is illuminance and the corresponding environmental control device 200 is a lighting device, the physical quantity prediction model 171 is represented by Equation (7) as described above.

[0116] The physical quantity prediction model 171 may be a linear regression model or a non-linear regression model. In this case, the parameter values of the model can be identified using learning data in the form of input-output data pairs. Examples of non-linear regression models include decision trees, support vector regression with non-linear kernels, neural networks, and the like. When identifying the parameter values, the values identified using learning data obtained in advance through experiments or the like can be used as initial values. Also, the physical quantity prediction model learning unit 188 may update the parameters. The parameter value identification algorithm may be executed in an appropriate manner according to the functional form of the model. For example, in the case of a linear regression model, the parameters can be identified by support vector regression. However, the configuration of the physical quantity prediction model 171 is not limited to a specific configuration and can be various configurations to which machine learning can be applied.

[0117] The arousal prediction model 172 is a mathematical model that can calculate the predicted change in the user's arousal level when a predetermined time has elapsed, based on the time average value and the time change amount of the physical quantity. When the physical quantities are temperature and illuminance, and the corresponding environmental control devices 200 are an air conditioner and a lighting device, respectively, the arousal prediction model is represented by Equations (8) to (10) as described above.

[0118] The arousal prediction model 172 may be a linear regression model or a non - linear regression model. In this case, the parameter values of the model can be identified using learning data in pairs of input - output data. Examples of non - linear regression models include decision trees, support vector regression with non - linear kernels, neural networks, etc. When identifying the parameter values, the values identified using learning data obtained in advance through experiments, etc., can be used as initial values. Also, the arousal prediction model learning unit 189 may update the parameters. The parameter value identification algorithm may be executed in an appropriate manner according to the functional form of the model. For example, in the case of a linear regression model, the parameters can be identified by support vector regression. However, the configuration of the arousal prediction model 172 is not limited to a specific configuration and can be various configurations to which machine learning is applicable.

[0119] Since the arousal optimization model is a non - linear discrete optimization problem, for example, it is solved by performing mathematical optimization calculations using meta - heuristic algorithms such as genetic algorithms or discrete PSO (Particle Swarm Optimization). Regarding the inequality constraint conditions (the above equations (4), (5), (12)), for example, the optimal solution can be calculated by converting to an unconstrained optimization problem using the penalty function method or using a meta - heuristic extended by a combination with the ε - constraint method, etc.

[0120] The numerical values of the constants and coefficients will be described. The value of the time step width Δτ is, for example, set to an appropriate value from the range of 15 to 30 minutes. From the viewpoints of the prediction accuracy of the arousal prediction model and the arousal effect, etc., 15 minutes is preferable for the value of the time step width Δτ. The time step index set T corresponds to the prediction horizon. In order to consider the stimulation of environmental changes due to time changes (such as temperature, cold, heat stimulation, etc.), the number of time steps needs to be 2 or more. From the balance with the computational amount, 3 or 4 is preferable for the number of time steps.

[0121] Lower limit value T of air-conditioning temperature setting value min , upper limit value T max , comfort value T best The values of may be set by the user by providing an input interface. The user may be allowed to input each of the three values. Only the comfort value T best may be input by the user, and the remaining two values may be set to 1°C before and after without sacrificing comfort, that is, "T min = T best - 1", "T max = T best + 1". Conversely, for the lower limit value T min , upper limit value T max , the user may be allowed to input them, and the remaining one value may be set to the average value, that is, "T best = (T min + T max ) / 2". By doing so, it is possible to calculate the air-conditioning temperature setting value at which the user can obtain an awakening effect within a warm and comfortable range.

[0122] Similarly, for the lower limit value L of the lighting output setting value min , upper limit value L max , comfort value L best The values of may be set by the user by providing an input interface. The user may be allowed to input each of the three values. Only the comfort value L best may be input by the user, and the remaining two values may be set to 20% before and after without sacrificing comfort, that is, "L min = L best - 20", "L max = L best + 20". Here, for the values of L min , L max , L best , percentage notation is used. For example, the value of L min may be set to 0%, and the value of L max may be set to 100%. Conversely, for the lower limit value L min , upper limit value L max , the user may be allowed to input them, and the remaining one value may be set to the average value, that is, "L best = (L min + L max) / 2」 may also be used. By doing so, it is possible to calculate the lighting output setting value that allows the user to obtain the awakening effect within a comfortable range in terms of brightness.

[0123] Penalty coefficient p of air conditioning temperature T is, as described above, the comfort value T best Since a range of ±1°C from this is the criterion for discomfort, "p T = 1 / 1 [point / °C]" is preferable. Penalty coefficient p of lighting output T is, as described above, the comfort value T best Since a range of ±20% from this is the criterion for discomfort, "p L = 1 / 20 [point / %]" is preferable.

[0124] Upper limit value p of penalty score max is preferably 1 or more and less than 2. For example, it is good to set "p max = 1.5". This is because it is empirically known that the tolerance for discomfort caused by a plurality of environmental control devices simultaneously is at most about two types. Here, the index set N of the subject and the arousal degree estimation value a i (τ) at the relative time τ are constants determined by the information acquired from the second acquisition unit. Therefore, it can be understood from the above that all constants and coefficients related to the calculation of the device setting value do not require special adjustment.

[0125] The calculation execution of the setting value calculation unit 184 is performed according to the procedure shown in FIG. 5 or FIG. 6. It is preferable to execute the calculation at regular intervals with a period of Δτ. FIG. 5 is a flowchart showing a first example of the procedure of the process in which the setting value calculation unit 184 calculates the device setting value and sets it in the environmental control device 200. FIG. 5 shows an example in which the setting value calculation unit 184 calculates the device setting value without using the arousal degree estimation value.

[0126] In the process of FIG. 5, the setting value calculation unit 184 determines whether the execution timing of the process of calculating the device setting value has arrived (step S100). If it is determined that the execution timing has not arrived (step S100: No), the process returns to step S100. As a result, the setting value calculation unit 184 waits for the arrival of the execution timing of the process of calculating the device setting value. On the other hand, when it is determined that the execution timing of the process of calculating the device setting value has arrived (step S100: Yes), the setting value calculation unit 184 acquires the device setting value from the monitoring control unit 181 (step S110).

[0127] Also, the setting value calculation unit 184 acquires the environmental measurement value (the measurement value of the physical quantity measured by the environmental measurement device 300) from the first acquisition unit 182 (step S120). Then, the setting value calculation unit 184 calculates the device setting value (the value for updating the device setting value) by solving the optimization problem as described above (step S130). In step S130, the setting value calculation unit 184 calculates the device setting value without using the estimated arousal value. The setting value calculation unit 184 outputs the obtained device setting value to the monitoring control unit 181 (step S140). The monitoring control unit 181 sets the device setting value obtained from the setting value calculation unit 184 in the environmental control device 200 by transmitting it to the environmental control device 200 via the communication unit 110. After step S140, the setting value calculation unit 184 ends the process of FIG. 5.

[0128] FIG. 6 is a flowchart showing a second example of the procedure of the process in which the setting value calculation unit 184 calculates the device setting value and sets it in the environmental control device 200. FIG. 6 shows an example in which the setting value calculation unit 184 calculates the device setting value using the estimated arousal value. Steps S200 to S220 in FIG. 6 are the same as steps S100 to S120 in FIG. 5. After step S220, the setting value calculation unit 184 acquires the estimated arousal value from the second acquisition unit 183 (step S230).

[0129] Then, the set value calculation unit 184 calculates a device set value (a value for updating the device set value) by solving the optimization problem as described above (step S240). In step S240, the set value calculation unit 184 calculates the device set value using the estimated arousal level. Step S250 is the same as step S140 in FIG. 5. After step S250, the set value calculation unit 184 ends the process of FIG. 6.

[0130] The calculation execution of the physical quantity prediction model learning unit 188 is performed according to the procedure shown in FIG. 7. For example, the calculation execution is performed at a fixed cycle, and the cycle ranges from 1 day to 2 weeks, with 1 day being preferred. FIG. 7 is a flowchart showing an example of the procedure of the process in which the physical quantity prediction model learning unit 188 calculates the parameter values of the physical quantity prediction model 171 by machine learning.

[0131] In the process of FIG. 7, the physical quantity prediction model learning unit 188 determines whether the execution timing of the process of calculating the parameter values has arrived (step S300). If it is determined that the execution timing has not arrived (step S300: No), the process returns to step S300. Thereby, the physical quantity prediction model learning unit 188 waits for the arrival of the execution timing of the process of calculating the parameter values. On the other hand, if it is determined that the execution timing of the process of calculating the parameter values has arrived (step S300: Yes), the physical quantity prediction model learning unit 188 determines whether the physical quantity prediction model 171 has not been learned (step S310).

[0132] Specifically, the physical quantity prediction model learning unit 188 attempts to acquire the parameters of the physical quantity prediction model 171 from the set value calculation unit 184. If the parameters can be acquired, the physical quantity prediction model learning unit 188 determines that the physical quantity prediction model 171 has been learned. On the other hand, if the parameters cannot be acquired, the physical quantity prediction model learning unit 188 determines that the physical quantity prediction model 171 has not been learned.

[0133] When it is determined that the physical quantity prediction model 171 has been learned (step S310: No), the physical quantity prediction model learning unit 188 determines whether a predetermined time has elapsed after learning (step S320). Specifically, the physical quantity prediction model learning unit 188 acquires the latest update date and time of the parameters of the physical quantity prediction model 171, and determines whether the difference between the latest update date and time and the current time exceeds the predetermined time by comparing them. The predetermined time in this case is, for example, a time within the range from 1 day to 2 weeks, and 1 week is preferable.

[0134] When it is determined that the predetermined time has not elapsed after learning (step S320: No), the physical quantity prediction model learning unit 188 acquires the device setting value from the monitoring control unit 181 (step S330). Also, the physical quantity prediction model learning unit 188 acquires the measured value (environment measurement value) of the physical quantity from the first acquisition unit 182 (step S340). Then, the physical quantity prediction model learning unit 188 evaluates the prediction accuracy of the physical quantity prediction model 171 using the acquired data and the parameter values set in the physical quantity prediction model 171, and determines whether the prediction accuracy of the physical quantity prediction model 171 has decreased (step S350).

[0135] For example, the physical quantity prediction model learning unit 188 sets the evaluation index of the prediction accuracy as the mean absolute error rate, or the correlation coefficient, etc., and determines whether the evaluation index is below a predetermined value. A plurality of evaluation indexes may be used for the determination. For example, the physical quantity prediction model learning unit 188 may determine that the prediction accuracy has decreased when both of the two evaluation indexes using the mean absolute error rate and the correlation coefficient are below the predetermined value.

[0136] When it is determined that the prediction accuracy of the physical quantity prediction model 171 has not decreased (step S350: No), the physical quantity prediction model learning unit 188 ends the process of FIG. 7. On the other hand, when it is determined that the prediction accuracy of the physical quantity prediction model 171 has decreased (step S350: Yes), the physical quantity prediction model learning unit 188 calculates the model parameter values by machine learning or the like using the acquired device setting values and the measured values of the physical quantity as learning data (step S360). The physical quantity prediction model learning unit 188 may execute machine learning or the like by an appropriate method according to the functional form of the physical quantity prediction model. For example, in the case of a linear regression model, the physical quantity prediction model learning unit 188 executes support vector regression.

[0137] The physical quantity prediction model learning unit 188 updates the parameter values of the physical quantity prediction model 171 by outputting the obtained parameter values to the setting value calculation unit 184 (step S370). When the physical quantity prediction model learning unit 188 evaluates the prediction accuracy using the parameters obtained by machine learning or the like, and the prediction accuracy is improved compared to before learning, the parameters and the date and time when the learning calculation was executed may be output to the setting value calculation unit 184. The evaluation index of the prediction accuracy may be the mean absolute error rate, the correlation coefficient, or the like. After step S370, the physical quantity prediction model learning unit 188 ends the process of FIG. 7.

[0138] On the other hand, in step S310, when it is determined that the physical quantity prediction model 171 is unlearned (step S310: Yes), the physical quantity prediction model learning unit 188 acquires the device setting values from the monitoring control unit 181 (step S331). Also, the physical quantity prediction model learning unit 188 acquires the measured values (environmental measurement values) of the physical quantity from the first acquisition unit 182 (step S341). After step S341, the process proceeds to step S360. On the other hand, in step S320, when it is determined that a predetermined time has elapsed after learning (step S320: Yes), the process proceeds to step S331.

[0139] The calculation execution of the arousal prediction model learning unit 189 is performed according to the procedure shown in FIG. 8. For example, the calculation execution is performed at regular intervals, and the interval ranges from one week to one month. An interval of one month is preferable. FIG. 8 is a flowchart showing an example of a procedure of a process in which the arousal prediction model learning unit 189 calculates parameter values of the arousal prediction model 172 by machine learning.

[0140] In the process of FIG. 8, the arousal prediction model learning unit 189 determines whether or not the execution timing of the process of calculating the parameter values has arrived (step S400). If it is determined that the execution timing has not arrived (step S400: No), the process returns to step S400. Thereby, the arousal prediction model learning unit 189 waits for the arrival of the execution timing of the process of calculating the parameter values. On the other hand, if it is determined that the execution timing of the process of calculating the parameter values has arrived (step S400: Yes), the arousal prediction model learning unit 189 determines whether the arousal prediction model 172 has not been learned (step S410).

[0141] Specifically, the arousal prediction model learning unit 189 attempts to acquire the parameters of the arousal prediction model 172 from the setting value calculation unit 184. If the parameters can be acquired, the arousal prediction model learning unit 189 determines that the arousal prediction model 172 has been learned. On the other hand, if the parameters cannot be acquired, the arousal prediction model learning unit 189 determines that the arousal prediction model 172 has not been learned.

[0142] If it is determined that the arousal prediction model 172 has been learned (step S410: No), the arousal prediction model learning unit 189 determines whether or not a predetermined time has elapsed after learning (step S420). Specifically, the arousal prediction model learning unit 189 acquires the latest update date and time of the parameters of the arousal prediction model 172, and determines whether the difference between the latest update date and time and the current time exceeds the predetermined time. This predetermined time is, for example, within a range of two weeks to six months, and two months is preferable.

[0143] When it is determined that a predetermined time has not elapsed after learning (step S420: No), the arousal level prediction model learning unit 189 acquires the arousal level estimated value from the second acquisition unit 183 (step S430). Further, the arousal level prediction model learning unit 189 acquires the measured value of the physical quantity (environmental measurement value) from the first acquisition unit 182 (step S440). Then, the arousal level prediction model learning unit 189 evaluates the prediction accuracy of the arousal level prediction model 172 using the acquired data and the parameter values set in the arousal level prediction model 172, and determines whether the prediction accuracy of the arousal level prediction model 172 has decreased (step S450).

[0144] For example, the arousal level prediction model learning unit 189 sets the evaluation index of the prediction accuracy as the mean absolute error rate, the correlation coefficient, etc., and determines whether the evaluation index is lower than a predetermined value. A plurality of evaluation indexes may be used in the determination. For example, when the arousal level prediction model learning unit 189 uses the mean absolute error rate and the correlation coefficient, it may be determined that the prediction accuracy has decreased when both of the two evaluation indexes are lower than the predetermined value.

[0145] When it is determined that the prediction accuracy of the arousal level prediction model 172 has not decreased (step S450: No), the arousal level prediction model learning unit 189 ends the process of FIG. 8. On the other hand, when it is determined that the prediction accuracy of the arousal level prediction model 172 has decreased (step S450: Yes), the arousal level prediction model learning unit 189 calculates the model parameter values by machine learning using the acquired arousal level estimated value and the measured value of the physical quantity (environmental measurement value) as learning data (step S460). The arousal level prediction model learning unit 189 may execute machine learning by an appropriate method according to the functional form of the physical quantity prediction model. For example, in the case of a linear regression model, the arousal level prediction model learning unit 189 executes support vector regression.

[0146] The arousal level prediction model learning unit 189 updates the parameter values of the arousal level prediction model 172 by outputting the obtained parameter values to the set value calculation unit 184 (step S470). The arousal prediction model learning unit 189 may evaluate the prediction accuracy using the parameters obtained by machine learning or the like, and output the parameters and the date and time when the learning calculation was executed to the set value calculation unit 184 when the prediction accuracy is improved compared to before learning. The evaluation index of the prediction accuracy may be the mean absolute error rate, the correlation coefficient, or the like. After step S470, the arousal prediction model learning unit 189 ends the process of FIG. 8.

[0147] On the other hand, when it is determined in step S410 that the arousal prediction model 172 is unlearned (step S410: Yes), the arousal prediction model learning unit 189 acquires the arousal estimation value from the second acquisition unit 183 (step S431). Further, the arousal prediction model learning unit 189 acquires the measured value (environmental measurement value) of the physical quantity from the first acquisition unit 182 (step S441). After step S441, the process proceeds to step S460. On the other hand, when it is determined in step S420 that a predetermined time has elapsed after learning (step S420: Yes), the process proceeds to step S431.

[0148] The set value determination unit 187 performs the process according to the procedure shown in FIGS. 9 to 12. For example, the calculation execution is performed at regular intervals, and the interval ranges from 1 day to 2 weeks. An interval of 1 day is preferable. FIG. 9 is a diagram showing a first example of the processing procedure for the set value determination unit 187 to determine and output the device set value. The set value determination unit 187 performs the process of FIG. 9 for each of the physical quantity prediction model 171 and the arousal prediction model 172. For example, the set value determination unit 187 may apply the process of FIG. 9 to each of the physical quantity prediction model 171 and the arousal prediction model 172 while taking synchronization, and in the conditional branch, make a branch by comprehensively considering the determination results of both models. In the process of Fig. 9, the setting value determination unit 187 determines whether the execution timing of the process for determining the device setting value has arrived (step S500). If it is determined that the execution timing has not arrived (step S500: No), the process returns to step S500. As a result, the setting value determination unit 187 waits for the arrival of the execution timing of the process for determining the device setting value.

[0149] On the other hand, if it is determined that the execution timing of the process for determining the device setting value has arrived (step S500; Yes), the setting value determination unit 187 determines whether any of the physical quantity prediction model 171 and the arousal level prediction model 172 that is the processing target is unlearned (step S510). Specifically, the setting value determination unit 187 attempts to obtain the parameters of either the physical quantity prediction model 171 or the arousal level prediction model 172 that is the processing target from the setting value calculation unit 184. The setting value determination unit 187 determines that the model for which the parameters could be obtained has been learned. On the other hand, the setting value determination unit 187 determines that the model for which the parameters could not be obtained is unlearned.

[0150] If it is determined that the model to be processed has been learned (step S510: No), the setting value determination unit 187 obtains the device setting value from the monitoring control unit 181 (step S520). Also, when the model to be processed is the physical quantity prediction model 171, the setting value determination unit 187 obtains the measured value (environmental measurement value) of the physical quantity from the first acquisition unit 182 (step S530). When the model to be processed is the arousal level prediction model 172, the setting value determination unit 187 obtains the estimated arousal level from the second acquisition unit 183 in step 530.

[0151] Then, the setting value determination unit 187 evaluates the prediction accuracy of the model to be processed using the acquired data and the parameter values of the model to be processed, and determines whether the prediction accuracy has decreased (step S540).

[0152] For example, the setting value determination unit 187 uses, as an evaluation index for prediction accuracy, the mean absolute error rate, the correlation coefficient, or the like, and determines whether or not the evaluation index is lower than a predetermined value. A plurality of evaluation indexes may be used for the determination. For example, the setting value determination unit 187 may be configured to determine that the prediction accuracy has decreased when both of the two evaluation indexes are lower than the predetermined values using the mean absolute error rate and the correlation coefficient.

[0153] When it is determined that the prediction accuracy of the model to be processed has not decreased (step S540: No), the setting value determination unit 187 ends the process of FIG. 9. On the other hand, when it is determined that the prediction accuracy of the model to be processed has decreased (step S540: Yes), the setting value determination unit 187 determines the device setting value (step S550). Specifically, the setting value determination unit 187 determines a device setting value that fluctuates as much as possible within the upper and lower limit ranges of the device setting value. The upper and lower limit ranges of the device setting value are shown in, for example, Expressions (4) and (5). For example, the setting value determination unit 187 may cause the device setting value to be periodically changed within the range between the upper limit value and the lower limit value as shown in FIG. 10.

[0154] FIG. 10 is a diagram showing an example of setting a device setting value by the setting value determination unit 187 for the environmental control device 200. Part (A) of FIG. 10 shows an example of setting a device setting value by the setting value determination unit 187 when the environmental control device 200 is an air conditioner. The horizontal axis of the graph in part (A) of FIG. 10 indicates time. The vertical axis indicates the device setting value (air conditioning setting value (air conditioning temperature setting value)). In the example of part (A) of FIG. 10, the setting value determination unit 187 greatly changes the air conditioning setting value within the range between its upper limit value and lower limit value. Specifically, the setting value determination unit 187 periodically changes the air conditioning setting value between its upper limit value and lower limit value. In this way, by greatly changing the device setting value of the environmental control device 200 (the air conditioner in the example of FIG. 10) by the setting value determination unit 187, data with a relatively large change in the explanatory variable value, which is suitable for machine learning or the like, can be obtained.

[0155] Part (B) of FIG. 10 shows an example of setting the device setting value by the setting value determination unit 187 when the environmental control device 200 is a lighting device. The horizontal axis of the graph in part (B) of FIG. 10 represents time. The vertical axis represents the device setting value (lighting setting value (lighting output setting value)). In the example of part (B) of FIG. 10, similar to the case of part (A) of FIG. 10, the setting value determination unit 187 greatly changes the lighting setting value within the range of its upper limit value and lower limit value. Also, when comparing the example of part (A) of FIG. 10 and the example of part (B) of FIG. 10, the setting value determination unit 187 changes the device setting value at different periods for the air conditioning device and the lighting device. Specifically, the setting value determination unit 187 changes the device setting value of the lighting device at a period that is half of that in the case of the air conditioning device.

[0156] In this way, when there are multiple types of environmental control devices 200, the setting value determination unit 187 changes the device setting value so that the combinations of the upper limit value and the lower limit value of the device setting values of the multiple types of environmental control devices 200 cover all cases with different periods. At this time, since the temporal change of the temperature is slower than the temporal change of the illuminance, it is preferable to set the variation period of the device setting value of the air conditioning device to be longer than the variation period of the device setting value of the lighting device. In addition, the setting value determination unit 187 may determine the device setting value according to a predetermined variation pattern, or may simply determine the device setting value randomly.

[0157] When the setting value determination unit 187 calculates the device setting value (step S550 in the example of FIG. 9), it transmits the execution non - permission information for the setting value calculation to the setting value calculation unit 184. The execution non - permission information is given an expiration date, and the setting value calculation unit 184 that has acquired the execution non - permission information does not execute a series of processes (steps S110 - S140 or S210 - S240) until the expiration date. For example, in step S100 or S200, the process branches to "No".

[0158] After step S550, the setting value determination unit 187 outputs the determined device setting value to the setting value calculation unit 184, thereby setting or updating the parameter value of the model to be processed (step S560). After step S560, the setting value determination unit 187 ends the process of FIG. 9. On the other hand, when it is determined in step S510 that the model to be processed is unlearned (step S510: Yes), the setting value determination unit 187 acquires the device setting value from the monitoring control unit 181 (step S521).

[0159] Also, when the model to be processed is the physical quantity prediction model 171, the setting value determination unit 187 acquires the measured value of the physical quantity from the first acquisition unit 182 (step S531). When the model to be processed is the arousal level prediction model 172, the setting value determination unit 187 acquires the estimated arousal level value from the second acquisition unit 183 in step 531. After step S531, the process proceeds to step S550.

[0160] FIG. 11 is a diagram showing a second example of the processing procedure for the setting value determination unit 187 to determine and output the device setting value. While FIG. 9 shows an example when the model to be processed is the physical quantity prediction model 171, FIG. 11 shows an example when the model to be processed is the arousal level prediction model 172. Therefore, in any of steps S520 and S521 of FIG. 9, the setting value determination unit 187 acquires the device setting value, whereas in any of steps S620 and S621 of FIG. 11, the setting value determination unit 187 acquires the estimated arousal level value. In other respects, the processing in FIG. 11 is the same as that in the case of FIG. 9.

[0161] As described above, when the setting value determination unit 187 calculates the device setting value, it transmits the execution non - permission information for the setting value calculation to the setting value calculation unit 184. In FIG. 11, the setting value determination unit 187 transmits the execution non - permission information for the setting value calculation to the setting value calculation unit 184 in step S650.

[0162] FIG. 12 is a diagram showing a third example of a processing procedure in which the setting value determination unit 187 determines and outputs device setting values. FIG. 9 shows an example when the model to be processed is the physical quantity prediction model 171, and FIG. 11 shows an example when the model to be processed is the arousal level prediction model 172. In contrast, FIG. 12 shows an example when both the physical quantity prediction model 171 and the arousal level prediction model 172 are the processing targets.

[0163] Therefore, in the processing of FIG. 9, the setting value determination unit 187 acquires the device setting values, and in the processing of FIG. 11, the setting value determination unit 187 acquires the arousal level estimated values. In contrast, in the processing of FIG. 12, the setting value determination unit 187 acquires both the device setting values and the arousal level estimated values. Specifically, the setting value determination unit 187 acquires the device setting values in steps S720 and S721 of FIG. 12, and acquires the arousal level setting values in steps S730 and S731.

[0164] In step S710 of FIG. 12, when the determination is Yes for either one of the physical quantity prediction model 171 and the arousal level prediction model 172, the processing may be branched to Yes. Regarding step S750 as well, when the determination is Yes for either one, the processing may be branched to Yes. Regarding step S700 as well, when the determination is Yes for either one, the processing may be branched to Yes. In this way, synchronization may be taken between the processing of the physical quantity prediction model 171 and the processing of the arousal level prediction model 172, and conditional branching may be performed by comprehensively considering both models. Alternatively, when the branch destinations are different between the physical quantity prediction model 171 and the arousal level prediction model 172, the processing of FIG. 12 may be performed separately. Regarding other points, the processing of FIG. 12 is the same as in the cases of FIG. 9 and FIG. 11.

[0165] As described above, when the setting value determination unit 187 calculates the device setting value, it transmits to the setting value calculation unit 184 information indicating non - permission to execute the setting value calculation. In FIG. 12, in step S760, the setting value determination unit 187 transmits to the setting value calculation unit 184 information indicating non - permission to execute the setting value calculation.

[0166] As described above, the physical quantity prediction model 171 calculates a predicted value of the physical quantity based on the setting value of the physical quantity that affects the arousal level of the subject. The arousal level prediction model 172 calculates a predicted value of the arousal level based on the predicted value of the physical quantity calculated by the physical quantity prediction model 171. The setting value calculation unit 184 uses the physical quantity prediction model 171 and the arousal level prediction model 172 to calculate a setting value (device setting value) for controlling the arousal level of the subject under the constraint conditions related to the physical quantity. The monitoring and control unit 181 sets the calculated setting value in the environmental control device 200.

[0167] In this way, by the physical quantity prediction model 171 calculating the predicted value of the physical quantity and the arousal level prediction model 172 predicting the arousal level using the predicted value of the physical quantity, the possibility of change in the physical quantity can be incorporated into the prediction of the arousal level. According to the arousal level control device 100, in this regard, when controlling the arousal level, the influence of the action on the surrounding environment on the arousal level can be grasped more accurately. Further, according to the arousal level control device 100, as described above, no special adjustment is required for all constants and coefficients related to the calculation of the device setting value.

[0168] Also, the setting value calculation unit 184 calculates the setting value so that the arousal level becomes higher. According to the arousal level control device 100, the arousal level of the subject can be improved. For example, when the subject is performing work, the work efficiency can be improved.

[0169] Also, the physical quantity prediction model 171 is a mathematical model that can calculate a predicted value of the physical quantity when a predetermined time has elapsed based on the measured value of the physical quantity and the setting value of the controlled device. Since the physical quantity prediction model 171 is configured as a mathematical model, it is possible to attach a meaning to the mathematical formulas and the like that constitute the physical quantity prediction model 171. By interpreting this meaning, the validity of the physical quantity prediction model 171 can be verified.

[0170] The arousal level prediction model 172 is a mathematical model that can calculate a predicted value of the change in the arousal level of a subject when a predetermined time has elapsed, based on the time average value and the change amount of a physical quantity. Since the arousal level prediction model 172 is configured as a mathematical model, it is possible to attach a meaning to the mathematical formulas and the like that constitute the arousal level prediction model 172. By interpreting this meaning, the validity of the arousal level prediction model 172 can be verified.

[0171] Also, the setting value calculation unit 184 calculates a setting value using the arousal level optimization model, which is a mathematical model, so that the value of the objective function becomes larger. Specifically, the setting value calculation unit 184 solves the optimization problem indicated by this arousal level optimization model. This arousal level optimization model includes, as constraint conditions, a first constraint condition that is a physical quantity prediction model, a second constraint condition that is an arousal level prediction model, and a third constraint condition that the setting value of the environmental control device is within a predetermined range. Further, the objective function of this arousal level optimization model is a function that calculates the total value or the average value of the predicted values of the change in the arousal level for one or more subjects and for one or more intervals of time steps. Since the setting value calculation unit 184 uses the arousal level optimization model, which is a mathematical model, it is possible to attach a meaning to the mathematical formulas and the like that constitute the arousal level optimization model. By interpreting this meaning, the validity of the arousal level optimization model can be verified. By verifying the validity of the arousal level optimization model, the validity of the processing performed by the setting value calculation unit 184 can be verified.

[0172] Also, the setting value calculation unit 184 calculates a setting value so that the trimmed average value of the predicted values of the change in the arousal level for one or more subjects and for one or more intervals of time steps becomes larger. By using the interrupt average in the set value calculation unit 184, for example, among the subjects, even if there are people with extremely small or extremely large changes in arousal level with respect to changes in physical quantities, these extreme subjects will not be over-evaluated, so overall optimization can be achieved.

[0173] Also, the set value calculation unit 184 calculates a set value that satisfies the constraint conditions regarding the comfort score calculated for the set value. Thereby, the set value calculation unit 184 can calculate a set value that takes into account not only the arousal level but also the comfort. In this regard, the arousal control device 100 can balance the arousal level and the comfort.

[0174] Also, the set value calculation unit 184 calculates each of a plurality of types of set values so that the total sum of the comfort penalty scores calculated for each of the plurality of types of set values is within a predetermined range. Thereby, when there are a plurality of types of environmental control devices 200, the set value calculation unit 184 can ensure comfort for the entire plurality of types of environmental control devices 200. According to the arousal control device 100, in this regard, comfort can be ensured, and the degree of freedom in setting the set value is greater than when ensuring comfort by controlling only one type of environmental control device 200.

[0175] Also, the arousal prediction model 172 calculates a predicted value of arousal based on the amount of change in the physical quantity in addition to the predicted value of the physical quantity. Arousal is considered to be sensitive to the magnitude of changes in physical quantities. By having the arousal prediction model 172 predict arousal based on the magnitude of the physical quantity, it is expected that arousal can be predicted with high accuracy.

[0176] Also, the arousal prediction model 172 calculates a predicted value of arousal based on at least the temporal variation of arousal. When the temporal variation (degree of variation) of the arousal level is large, it is considered that the subject is drowsy, and the arousal level of the subject is considered to be relatively low. By predicting the arousal level based on the temporal variation of the arousal level by the arousal level prediction model 172, it is expected that the current state of the arousal level can be grasped more accurately, and it is expected that the prediction accuracy of the arousal level will be improved.

[0177] In addition, the physical quantity prediction model learning unit 188 performs machine learning or the like based on the measured value of the physical quantity and the set value of the physical quantity, and obtains the set parameter value of the physical quantity prediction model 171. Thereby, the physical quantity prediction model learning unit 188 can automatically obtain the set parameter value of the physical quantity prediction model 171, and there is no need to obtain the set parameter value manually by the administrator of the arousal level control device 100 or the like. According to the arousal level control device 100, in this regard, the labor for setting the physical quantity prediction model 171 can be reduced.

[0178] In addition, the physical quantity prediction model learning unit 188 performs machine learning or the like when at least one of the following cases occurs: when the set parameter value of the physical quantity prediction model 171 is not set, when the prediction accuracy by the physical quantity prediction model 171 drops below a predetermined condition, and when a predetermined time or more has elapsed since the setting of the set parameter value of the physical quantity prediction model 171. Thereby, the physical quantity prediction model learning unit 188 can perform machine learning or the like as needed. Therefore, for example, the processing load of the physical quantity prediction model learning unit 188 can be lighter than when machine learning or the like is repeatedly performed constantly.

[0179] In addition, the arousal level prediction model learning unit 189 performs machine learning based on the measured value of the physical quantity and the arousal level, and obtains the set parameter value of the arousal level prediction model. Thereby, the arousal level prediction model learning unit 189 can automatically obtain the set parameter value of the arousal level prediction model 172. Therefore, there is no need to obtain the set parameter value manually by the administrator of the arousal level control device 100 or the like. According to the arousal level control device 100, in this regard, the labor for setting the arousal level prediction model 172 can be reduced.

[0180] Further, when the set parameter value of the arousal level prediction model 172 is not set, when the prediction accuracy by the arousal level prediction model 172 drops below a predetermined condition, and when at least one of the cases where a predetermined time or more has elapsed since the setting of the set parameter value of the arousal level prediction model 172 occurs, the arousal level prediction model learning unit 189 performs machine learning. Thereby, the arousal level prediction model learning unit 189 can perform machine learning or the like as necessary. Therefore, for example, the processing load on the arousal level prediction model learning unit 189 can be reduced as compared with the case of constantly repeating machine learning or the like.

[0181] Also, the set value determination unit 187 determines the set value within a predetermined range of the set value. When the set parameter value of the physical quantity prediction model 171 is not set, when the prediction accuracy by the physical quantity prediction model 171 drops below a predetermined condition, when a predetermined time or more has elapsed since the setting of the set parameter value of the physical quantity prediction model 171, when the set parameter value of the arousal level prediction model 172 is not set, when the prediction accuracy by the arousal level prediction model 172 drops below a predetermined condition, and when at least one of the cases where a predetermined time or more has elapsed since the setting of the set parameter value of the arousal level prediction model 172 occurs, the set value determination unit 187 sets the set value determined by itself in the environment control device 200 instead of the set value calculated by the set value calculation unit 184.

[0182] By determining the set value, the set value determination unit 187 can set a set value other than the optimal solution in the arousal level control by the set value calculation unit 184 in the environment control device 200, and obtain a wider range of learning data. At least one of the physical quantity prediction model learning unit 188 and the arousal level prediction model learning unit 189 performs machine learning or the like using the learning data obtained based on the set value determined by the set value determination unit 187, so that machine learning can be performed with higher accuracy, and in this regard, the accuracy of the model can be improved.

[0183] Next, the configuration of the embodiment of the present invention will be described with reference to FIG. 13. FIG. 13 is a diagram showing an example of the configuration of the arousal level control device 10 according to the embodiment. The arousal level control device 10 shown in FIG. 13 includes a physical quantity prediction model (a storage unit storing the physical quantity prediction model) 11, an arousal level prediction model (a storage unit storing the arousal level prediction model) 12, a set value calculation unit 13, and a setting unit 14. With such a configuration, the physical quantity prediction model 11 calculates a predicted value of the physical quantity based on the set value of the physical quantity that affects the arousal level of the subject. The arousal level prediction model 12 calculates a predicted value of the arousal level based on the predicted value. The set value calculation unit 13 calculates the set value for controlling the arousal level of the subject under the constraint conditions regarding the physical quantity, using the physical quantity prediction model and the arousal level prediction model. The setting unit 14 sets the calculated set value in the control target device that affects the physical quantity.

[0184] In this way, by having the physical quantity prediction model 11 calculate the predicted value of the physical quantity and the arousal level prediction model 12 predict the arousal level using the predicted value of the physical quantity, it is possible to incorporate the anticipation of changes in the physical quantity into the prediction of the arousal level. According to the arousal level control device 10, in this regard, it is possible to more accurately grasp the influence of the action on the surrounding environment on the arousal level during arousal level control.

[0185] The configuration of the arousal level control device 100 is not limited to a configuration using a computer. For example, the arousal level control device 100 may be configured using dedicated hardware such as being configured using an ASIC (Application Specific Integrated Circuit).

[0186] An embodiment of the present invention can also be realized by causing a CPU (Central Processing Unit) to execute a computer program for any processing. In this case, the program can be stored using various types of non-transitory computer readable media and supplied to a computer. The non-transitory computer readable media includes various types of tangible storage media. Examples of non-transitory computer readable media include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROM (Read Only Memory), CD-R, CD-R / W, DVD (Digital Versatile Disc), BD (Blu-ray (registered trademark) Disc), and semiconductor memories (e.g., mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)).

[0187] This application claims priority based on Japanese Patent Application No. 2019-018212 filed on February 4, 2019, and incorporates the entire disclosure thereof herein.

Industrial Applicability

[0188] The present invention may be applied to an arousal control device, an arousal control method, and a recording medium.

Explanation of Signs

[0189] 1 Arousal control system 10, 100 Arousal control device 11, 171 Physical quantity prediction model 12, 172 Arousal prediction model 13, 184 Set value calculation unit (set value calculation means) 14 Setting unit (setting means) 110 Communication unit (communication means) 170 Storage unit (storage means) 180 Control unit (control means) 181 Monitoring control unit (monitoring control means) 182 First acquisition unit (first acquisition means) 183 Second acquisition unit (second acquisition means) 185 Physical quantity prediction model calculation unit (physical quantity prediction model calculation means) 186 Arousal level prediction model calculation unit (arousal level prediction model calculation means) 187 Set value determination unit (set value determination means) 188 Physical quantity prediction model learning unit (physical quantity prediction model learning means) 189 Arousal level prediction model learning unit (arousal level prediction model learning means) 200 Environmental control device 300 Environmental measurement device 400 Arousal level estimation device

Claims

1. Within the range of the constraint conditions for the set value, where the set value set as the control target value of a physical quantity that affects at least any one of the temperature, illuminance, sound volume, and vibration magnitude of the surrounding environment of the subject of arousal control, or a combination thereof, is included in a predetermined range set by the user, a physical quantity prediction model, which is a model including the physical quantity and the set value as explanatory variables, and is a positive function represented by using, as the explained variable, the predicted value of the physical quantity after a lapse of a predetermined time from the timing when the physical quantity has become the measured value by the environmental measurement device that measures the physical quantity in the state where the set value is set in the control device, and the model parameters are identified by machine learning; and a model including, as explanatory variables, the physical quantity and its time change amount and the estimated value of the arousal level, and is a positive function represented by using, as the explained variable, the predicted value of the time change amount of the arousal level, and the model parameters are identified by machine learning. The objective function value of the objective function, which is a function for calculating the total value or average value of the predicted values of the time change amounts of the arousal levels of one or more of the subjects over one or more intervals of time steps, becomes larger, a set value calculation means for performing a solution search of the set value; a setting means for setting the calculated set value in the control device; An arousal control device comprising the above.

2. The constraint condition includes a comfort condition, which is a constraint condition that the weighted sum, for a plurality of physical quantities, of the comfort penalty scores calculated based on the magnitude of the difference between the set value and the comfort value defined as a comfortable physical quantity, with weights set according to the degree of influence of each of the plurality of physical quantities on the comfort of the subject, is included in a predetermined range. The arousal control device according to Claim 1.

3. The arousal prediction model includes, as an explanatory variable, the time change amount of the arousal level at a time prior to the time for which the time change amount of the arousal level is to be predicted. The arousal control device according to Claim 1 or Claim 2.

4. The arousal prediction model includes, as an explanatory variable, the temporal variation of the arousal level. The arousal control device according to any one of Claims 1 to 3.

5. By a computer Within the range of the constraint conditions for the set value, where the set value set as the control target value of the physical quantity that affects at least one of the temperature, illuminance, sound volume, and vibration magnitude of the surrounding environment of the subject of arousal control, or a combination thereof, is included in a predetermined range set by the user, a physical quantity prediction model that is a model including the physical quantity and the set value as explanatory variables, and is a positive function represented by using, as the explained variable, the predicted value of the physical quantity after a lapse of a predetermined time from the timing when the physical quantity has become the measured value by the environmental measurement device that measures the physical quantity in a state where the set value is set in the control device, and is a physical quantity prediction model obtained by identifying model parameters by machine learning, and a model including, as explanatory variables, the physical quantity and its time change amount and the estimated value of the arousal level, and is a positive function represented by using, as the explained variable, the predicted value of the time change amount of the arousal level, and is a model obtained by identifying model parameters by machine learning, and performing solution search of the set value so that the objective function value of the objective function, which is a function for calculating the total value or average value of the predicted values of the time change amount of the arousal level of the subject for one or more of the subjects and for at least one section of time steps, becomes larger, setting the calculated set value in the control device A method for controlling arousal level including this.

6. On a computer, Within the range of the constraint conditions for the set value, where the set value set as the control target value of the physical quantity that affects at least one of the temperature, illuminance, sound volume, and vibration magnitude of the surrounding environment of the subject of arousal control, or a combination thereof, is included in a predetermined range set by the user, A physical quantity prediction model which is a model including the physical quantity and the set value as explanatory variables, and is a positive function represented by using, as the explained variable, a predicted value of the physical quantity after a lapse of a predetermined time from the timing when the physical quantity has become a measured value by an environmental measurement device that measures the physical quantity, in a state where the set value is set in the control device, and the model parameters are identified by machine learning; and a model including, as explanatory variables, the physical quantity and its amount of change over time, and an estimated value of the arousal level, and is a positive function represented by using, as the explained variable, a predicted value of the amount of change over time of the arousal level, and the arousal level prediction model obtained by identifying the model parameters by machine learning as part of the function, and performing solution search of the set value so that the objective function value of an objective function, which is a function for calculating a sum value or an average value of predicted values of the amount of change over time of the arousal level of the subject for one or more of the subjects and for one or more intervals of time steps, becomes larger. Setting the calculated set value in the control device. A program for causing the above to be executed.

Images