WO2015037165A1 - 情報処理装置、予測制御方法及び記録媒体 - Google Patents
情報処理装置、予測制御方法及び記録媒体 Download PDFInfo
- Publication number
- WO2015037165A1 WO2015037165A1 PCT/JP2014/002441 JP2014002441W WO2015037165A1 WO 2015037165 A1 WO2015037165 A1 WO 2015037165A1 JP 2014002441 W JP2014002441 W JP 2014002441W WO 2015037165 A1 WO2015037165 A1 WO 2015037165A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- control
- operation amount
- information
- prediction
- control target
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
Definitions
- the present invention relates to control of processes and the like, and more particularly, to predictive control for predicting a future state and determining an operation amount.
- Predictive control is control that predicts the future controlled quantity based on the dynamic model of the process and determines the manipulated variable.
- the features of predictive control are that multivariable control is easy, constraints are easily considered, and adjustment is easy and intuitive.
- Predictive control has been used mainly for petrochemical industry plant control. Along with the recent improvement in computer capabilities, predictive control is applied not only to plant control in the petrochemical industry but also to mobile objects and robots with short control cycles (see, for example, Non-Patent Document 1).
- Non-Patent Document 1 General predictive control including the technology described in Non-Patent Document 1 is based on the assumption that a predictive control model reflecting the real world is known and the predictive control model designer describes the predictive control model in advance. (For example, refer to Patent Document 1).
- the model predictive control apparatus described in Patent Literature 1 has a known predictive control model, and performs control to follow the model based on an evaluation function. Note that the predictive control model is often expressed as a discrete-time linear function.
- Patent Document 2 a control device that learns a predictive control model for control using a neural network is used (see, for example, Patent Document 2).
- the control device described in Patent Literature 2 includes an identifier that obtains parameters of an identification model (corresponding to the predictive control model described above).
- the control device described in Patent Document 2 is also based on the assumption that the predictive control model is known.
- Patent Document 3 a method for learning the parameters of the predictive control model has been proposed (see, for example, Patent Document 3).
- the method described in Patent Document 3 also has a known predictive control model, and estimates parameters based on the premise.
- Patent Document 4 a system that operates by selecting a plurality of fixed control algorithms to match individual differences has been proposed (see, for example, Patent Document 4).
- Patent Document 5 An apparatus using machine learning has been proposed (see, for example, Patent Document 5).
- model predictive control technology described in Non-Patent Document 1 has a predictive control model. It is assumed that it is already known.
- the predictive control designer described the predictive control model in advance based on physical laws such as the equation of motion, heat equation, mass conservation law, momentum conservation law, energy conservation law.
- a plurality of operation variables such as an air conditioner temperature, an air conditioner air volume, a blind opening, and a window opening affect a plurality of control variables such as a room temperature and a room humidity.
- control variables such as increasing temperature and decreasing humidity.
- state variables such as outside air temperature and outside air humidity affect the control variables.
- the predictive control has a problem that the followability of the control is low.
- the predictive control model is greatly increased from the real world over time based on changes in the surrounding environment, changes in the performance of the control target over time, or structural changes in the control target based on modification of the control target. Separate. For this reason, the predictive control model of the predictive technique is an incomplete predictive control model. As a result, the predictive control has a problem that the followability of the control is lowered.
- control using machine learning the control characteristics change every learning, so it is necessary to evaluate the characteristics of the control model each time. And it is determined whether it is appropriate control based on the evaluation. For this reason, the control using machine learning described in the literature has a problem that it is difficult to select an appropriate control based on the surrounding situation.
- An object of the present invention is to provide an information processing apparatus and a predictive control method that solve the above problems.
- An information processing apparatus includes an information storage unit that receives and stores control target information including information on a control target and a surrounding environment including the control target, and control target information stored in the information storage unit.
- a predictive expression set learning generating means for learning and generating a predictive expression set used for determining the manipulated variable of the controlled object based on the input information necessary for determining the manipulated variable of the controlled object, and the predictive expression set And the control target information stored in the information storage means, the received control target information, and the input information, constructing a predictive control model of the control target, and an operation amount used for control of the control target And an operation amount determining means for determining.
- a predictive control method receives control target information including information on a control target and surrounding environment including the control target, accumulates the control target information, and based on the accumulated control target information Learning and generating a prediction formula set used for determining the operation amount of the control target, receiving input information necessary for determining the operation amount of the control target, the prediction formula set, the accumulated control target information, and the Based on the received control target information and the input information, a predictive control model of the control target is constructed, and an operation amount that uses the control target for control is determined.
- a computer-readable recording medium on which a program according to an aspect of the present invention is recorded is a process of receiving control target information including information on a control target and surrounding environment including the control target, and storing the control target information; A process for learning and generating a prediction equation set used for determining the operation amount of the control object based on the accumulated control object information, receiving input information necessary for determining the operation amount of the control object, and receiving the prediction equation A process of constructing a predictive control model of the control target based on the set, the accumulated control target information, the received control target information, and the input information, and determining an operation amount to use the control target for control And let the computer run.
- An information processing apparatus includes an information storage unit that receives and stores control target information including information on a control target and a surrounding environment including the control target, and control target information stored in the information storage unit.
- a predictive equation set learning generating means for learning and generating a predictive equation set used for determining an operation amount for the control object based on the input information, and receiving input information necessary for determining the operation amount of the control object;
- the control target information stored in the information storage means and the prediction formula set, a predictive control model of the control target is constructed, and a first operation amount used for control of the control target is determined.
- An operation amount determination unit, and a control unit that outputs the first operation amount, a control formula that is input in advance, and control target information that is stored in the information storage unit.
- Fixed control means for determining and outputting a second manipulated variable used in the operation, and an manipulated variable for selecting one of the first manipulated variable of the manipulated variable determiner and the second manipulated variable of the fixed control means Selecting means.
- the predictive control method receives and accumulates control target information including information related to a control target and a surrounding environment including the control target, and performs control on the control target based on the stored control target information. Learning and generating a prediction formula set used for determining an operation amount, receiving input information necessary for determining an operation amount of the control target, and receiving the input information, the accumulated control target information, and the prediction formula set.
- a predictive control model of the control object is constructed, a first operation amount used for the control of the control object is determined, the first operation amount is output, and a formula inputted in advance and the accumulation Based on the controlled object information, a second operation amount used for controlling the control object is determined and output, and one of the first operation amount and the second operation amount is selected.
- a computer-readable recording medium on which a program according to an aspect of the present invention is recorded is a process of receiving and storing control target information including information on a control target and surrounding environment including the control target, and the stored control target
- learning data can be collected and controlled simultaneously.
- FIG. 1 is a block diagram showing an example of the configuration of an information processing system including an information processing apparatus according to the first embodiment of the present invention.
- FIG. 2 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the first embodiment.
- FIG. 3 is a block diagram illustrating an example of another configuration of the information processing system including the information processing apparatus according to the first embodiment.
- FIG. 4 is a block diagram illustrating an example of another configuration of the information processing system including the information processing apparatus according to the first embodiment.
- FIG. 5 is a block diagram illustrating an exemplary configuration of a modification of the information processing apparatus according to the first embodiment.
- FIG. 6 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the second embodiment.
- FIG. 1 is a block diagram showing an example of the configuration of an information processing system including an information processing apparatus according to the first embodiment of the present invention.
- FIG. 2 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the first embodiment.
- FIG. 7 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the third embodiment.
- FIG. 8 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the fourth embodiment.
- FIG. 9 is a block diagram illustrating an example of a configuration of an operation amount determination unit according to the fifth embodiment.
- FIG. 10 is a block diagram illustrating an example of a configuration of an operation amount calculation unit according to the fifth embodiment.
- FIG. 11 is a block diagram illustrating an example of another configuration of the operation amount calculation unit according to the fifth embodiment.
- FIG. 12 is a block diagram illustrating an example of a configuration of a mathematical programming problem formulation unit according to the fifth embodiment.
- FIG. 13 is a diagram for explaining formulation in the mathematical programming problem formulation unit and calculation in the mathematical programming problem calculation unit.
- FIG. 13 is a diagram for explaining formulation in the mathematical programming problem formulation unit and calculation in the mathematical programming problem calculation unit.
- FIG. 14 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the sixth embodiment.
- FIG. 15 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the sixth embodiment.
- FIG. 16 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the sixth embodiment.
- FIG. 17 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the sixth embodiment.
- FIG. 18 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the sixth embodiment.
- FIG. 19 is a block diagram illustrating an example of a configuration of an information processing device according to the sixth embodiment.
- FIG. 20 is a block diagram illustrating an example of the configuration of the information processing apparatus according to the sixth embodiment.
- FIG. 1 is a block diagram illustrating an example of a configuration of an information processing system 100 including the information processing apparatus 10000 according to the first embodiment of this invention.
- the information processing system 100 includes an information processing device 10000, an input device 22000, and an output device 23000.
- the information processing apparatus 10000 executes processing related to predictive control for the control target 21000 included in the surrounding environment 20000. Therefore, the information processing device 10000 receives information on the surrounding environment 20000 and the control target 21000. However, each information is subject to various disturbances when measuring the value. Here, the disturbance includes “measuring disturbance” that can be measured and “unmeasured disturbance” that cannot be measured. Therefore, the information processing device 10000 receives the following information as information related to the control target 21000 and the surrounding environment 20000 as shown in FIG.
- the information processing device 10000 includes information obtained by adding measurement disturbance and unmeasured disturbance to information on the surrounding environment 20000, information obtained by adding measurement disturbance and unmeasured disturbance to information on the control target 21000, and measurement disturbance (hereinafter, Collectively referred to as “control target information 2100”).
- the information of the control target 21000 is, for example, the state of the control target 21000, an operation applied to the control target 21000, or a cost for realizing the operation.
- the information processing device 10000 is connected to the input device 22000 and receives information necessary for predictive control (for example, setting values, constraints, priority.
- information necessary for predictive control for example, setting values, constraints, priority.
- input information 2200 simply referred to as “input information 2200”
- the input device 22000 is not particularly limited as long as the designer can input necessary information.
- the input device 22000 may be a general PC (Personal Computer). Therefore, detailed description of the input device 22000 is omitted.
- the information processing device 10000 is connected to the output device 23000 and outputs information for operating the determined control target 21000 (hereinafter simply referred to as “operation amount 2300”) to the output device 23000.
- the output device 23000 is not particularly limited as long as the designer can confirm the operation amount 2300.
- the output device 23000 may be a general display device. Therefore, detailed description of the output device 23000 is omitted.
- the input device 22000 and the output device 23000 may be different devices or the same device.
- FIG. 2 is a block diagram illustrating an example of the configuration of the information processing apparatus 10000 according to the first embodiment.
- the information processing apparatus 10000 includes an information storage unit 11000, a prediction formula set learning generation unit 12000, and an operation amount determination unit 14000.
- the information storage unit 11000 receives the control target information 2100 and stores it.
- the information accumulated by the information accumulation unit 11000 is referred to as “accumulated information 2600”.
- the prediction formula set learning generation unit 12000 learns the prediction formula set 2500 used for controlling the control target 21000 based on the storage information 2600 stored in the information storage unit 11000, that is, the control target information 2100 from the past to the present ( Machine learning). And the prediction formula set learning production
- the prediction formula set 2500 is a set of prediction formulas for predicting the future state or cost of the control target 21000, for example.
- the operation amount determination unit 14000 receives information necessary for determining the operation amount 2300, that is, information necessary for predictive control (the input information 2200 described above). For example, the operation amount determination unit 14000 may receive the input information 2200 necessary for determining the operation amount 2300 from the input device 22000 as described above.
- the input information 2200 is, for example, a set value, constraints, and priority.
- the set value is a set of values set in a future point in time when the control object 21000 is in control and a variable (hereinafter referred to as a control variable) to be controlled by the predictive control model at the future point in time. At that time, control is performed so that the difference between the value of the control variable and the value set for the control variable to be controlled becomes small.
- the constraints are constraints on a control variable of the predictive control model of the control target 21000, a variable indicating the state of the predictive control model (hereinafter referred to as a state variable), a variable indicating an operation of the predictive control model (hereinafter referred to as an operation variable), It is a constraint among control variables, state variables, and manipulated variables.
- prediction formula can be said to be a function (function formula) having control variables, state variables, and operation variables as variables.
- the priority represents the importance of the control variable of the predictive control model of the control target 21000.
- a control variable having a high priority is preferentially controlled.
- the operation amount determination unit 14000 constructs a predictive control model based on the following information, and determines the operation amount 2300 of the control target 21000 based on the predictive control model. That is, the information includes the prediction formula set 2500 learned and generated by the prediction formula set learning generation unit 12000, the received control target information 2100, the accumulated information 2600 accumulated in the information accumulation unit 11000, and the input information 2200. It is.
- the operation amount 2300 is information used for controlling the control target 21000.
- the operation amount 2300 is, for example, a specific value of the operation. However, it is assumed that the operation amount that is not a specific value such as an operation instruction of the operation is information appropriately replaced with a specific value.
- the operation amount determination unit 14000 outputs the determined operation amount 2300 to the output device 23000.
- the operator of the control target 21000 can appropriately operate the control target 21000 based on the output information (operation amount 2300).
- the manipulated variable determination unit 14000 includes the independent variables and the dependent variables of the prediction formula set 2500 obtained based on the machine learning in the prediction formula set learning generation unit 12000, the control variables, the state variables, and the operation variables of the prediction control model. Specify the correspondence with. Then, the operation amount determination unit 14000 builds a prediction control model for control using the prediction formula set 2500 based on the definition. Furthermore, the information processing device 10000 may construct a predictive control model using the input information 2200 (for example, setting value, constraint, priority) received by the operation amount determination unit 14000. That is, the information processing apparatus 10000 can construct a predictive control model using the predictive formula set 2500 obtained based on machine learning.
- the input information 2200 for example, setting value, constraint, priority
- the operation of the operation amount determination unit 14000 including independent variables and dependent variables will be described later.
- the designer elucidates the relationship between the control object 21000, its surrounding environment 20000, and measurement disturbances, establishes a prediction formula in consideration of physical laws, and constructs a prediction control model.
- the prediction technique made the parameter of the prediction formula undecided and identified the parameter based on learning.
- the information processing apparatus 10000 can obtain an effect that enables a designer to perform predictive control without constructing a predictive control model.
- the prediction formula set learning generation unit 12000 automatically learns and generates the prediction formula set 2500 based on the accumulated information 2600 accumulated in the information accumulation unit 11000. To do.
- the operation amount determination unit 14000 automatically constructs a predictive control model and determines the operation amount 2300 based on the control target information 2100, the input information 2200, the accumulated information 2600, and the prediction formula set 2500. is there.
- the information processing apparatus 10000 can obtain an effect that enables predictive control without a designer constructing a predictive control model.
- the information processing apparatus 10000 can construct a predictive control model using a predictive expression set 2500 obtained based on machine learning.
- the predictive control described in the literature includes independent variables and dependent variables of prediction formulas obtained by simply applying machine learning, and operational variables, control variables, and control variables used in predictive control models for control. No association with state variables is specified. Therefore, the prediction technology cannot construct a predictive control model for control from the obtained prediction formula, and cannot construct a predictive control model for control by giving external setting values, constraints, priorities, etc. There was a problem.
- the information processing apparatus 10000 includes an independent variable and a dependent variable of the prediction formula set 2500 obtained based on the machine learning of the prediction formula set learning generation unit 12000, a control variable of the control model, and Specify the association with the manipulated variable. This is because the information processing apparatus 10000 constructs a predictive control model for control using the predictive formula set 2500 based on the rules. Furthermore, the information processing apparatus 10000 uses the input information 2200 received by the operation amount determination unit 14000. Thereby, the above-mentioned problem is solved.
- the information processing apparatus 10000 can perform predictive control for a control object 21000 that cannot be derived by a designer or a phenomenon that is not based on a physical law.
- the prediction formula set learning generation unit 12000 of the information processing apparatus 10000 generates a prediction formula set 2500 based on the accumulated information 2600.
- the operation amount determination unit 14000 can determine the operation amount 2300 based on the control target information 2100, the accumulated information 2600, and the prediction formula set 2500. That is, the information processing apparatus 10000 can determine the operation amount 2300 without using a physical law.
- the information processing apparatus 10000 can perform predictive control corresponding to changes in the surrounding environment 20000 and the control target 21000.
- the prediction formula set learning generation unit 12000 and the operation amount determination unit 14000 use the accumulated information 2600 that is the past control target information 2100 in addition to the control target information 2100. That is, in the information processing device 10000, the predictive formula set learning generation unit 12000 uses the machine learning to generate the predictive control model (predictive formula set 2500) so as to reduce the separation between the predictive control model and the real world. Build automatically. This is because the operation amount determination unit 14000 determines the operation amount 2300 of the prediction control using the constructed prediction control model (prediction formula set 2500).
- the information processing apparatus 10000 can realize prediction control using a prediction formula.
- prediction formula set learning generation unit 12000 learns and generates the prediction formula set 2500, and the operation amount determination unit 14000 uses the learned and generated prediction formula set 2500.
- the information processing apparatus 10000 of the present embodiment can obtain an effect of improving the followability of the predictive control.
- the number of control objects 21000 is not limited.
- FIG. 3 is a block diagram showing an example of the configuration of the information processing system 101 according to this modification.
- the information processing system 101 includes a plurality of control objects 21001 or control objects 2100n as control objects 21000.
- the information processing apparatus 10000 may be provided for each control target 21000.
- the information processing apparatus 10000 can operate even when there are a plurality of control objects 21000.
- the information processing apparatus 10000 receives the control target information 2100 related to the plurality of control targets 21000 and determines the operation amount 2300 of the plurality of control targets 21000.
- the information processing apparatus 10000 can centrally control the plurality of control objects 21000. That is, the information processing apparatus 10000 can easily realize control of the plurality of control objects 21000.
- the information processing apparatus 10000 can improve the prediction accuracy of the prediction formula set 2500 generated by the prediction formula set learning generation unit 12000.
- the information processing apparatus 10000 according to the present modification can control a plurality of control objects 21000 with the quantity of one control apparatus, a low quantity of the control apparatus can be realized.
- the information processing apparatus 10000 of the present modification can obtain an effect of simplifying the control of the plurality of control objects 21000 in addition to the effect of the information processing apparatus 10000 of the first embodiment.
- the reason is that the information processing apparatus 10000 can control a plurality of control objects 21000 in a centralized manner.
- the information processing apparatus 10000 can improve the prediction accuracy of the prediction formula set 2500.
- the reason is that the information processing apparatus 10000 can accumulate more information about the control target 21000.
- the information processing apparatus 10000 can obtain the effect of reducing the amount of material required for control.
- the reason is that the information processing apparatus 10000 can reduce the number of apparatuses required for the control target 21000.
- the information processing system 100 may transmit the operation amount 2300 determined by the information processing apparatus 10000 to the control target 21000.
- FIG. 4 is a block diagram showing an example of the configuration of the information processing system 102 according to this modification.
- the information processing apparatus 10000 transmits the operation amount 2300 determined by the operation amount determination unit 14000 to the control target 21000.
- the control target 21000 performs an operation based on the received operation amount 2300.
- the information processing apparatus 10000 may automatically feed back the operation amount.
- the information processing apparatus 10000 according to this modification has an effect of reducing the load on the operator in addition to the effect of the first embodiment.
- the reason is that the information processing apparatus 10000 can automatically feed back the determined operation amount 2300 to the control target 21000.
- the information processing apparatus 10000 of the present modification may output the operation amount 2300 of the output device 23000 in the same manner as in the first embodiment so that the operator can check the operation amount 2300.
- the information processing apparatus 10000 according to the present modification may be connected to a plurality of control objects 21000 as in the first modification.
- the configuration of the information processing device 10000 is not limited to the above description.
- the information processing apparatus 10000 may divide each configuration into a plurality of configurations.
- the information processing apparatus 10000 does not need to be configured by one apparatus.
- the information processing apparatus 10000 may use an external storage device connected via a network or a bus as the information storage unit 11000.
- At least a part of the configuration may be installed at a location different from the control target 21000.
- the information storage unit 11000 and the predictive set learning generation unit 12000 of the information processing device 10000 may be installed in a building different from the building where the control target 21000 is installed.
- the information storage unit 11000 and the predictive set learning generation unit 12000 of the information processing device 10000 may be installed in a country different from the country in which the control target 21000 is installed.
- the information processing apparatus 10000 may be connected to a plurality of control objects 21000 installed in a distributed manner.
- the information processing apparatus 10000 may have a plurality of configurations as one configuration.
- the information processing device 10000 may be realized as a computer device including a CPU (Central Processing Unit), a ROM (Read Only Memory), and a RAM (Random Access Memory).
- the information processing apparatus 10000 may be realized as a computer apparatus including an input / output connection circuit (IOC: Input ⁇ ⁇ Output Circuit) and a network interface circuit (NIC: Network Interface Circuit) in addition to the above configuration.
- IOC Input ⁇ ⁇ Output Circuit
- NIC Network Interface Circuit
- FIG. 5 is a block diagram illustrating an example of the configuration of the information processing apparatus 600 according to the third modification.
- the information processing apparatus 600 includes a CPU 610, a ROM 620, a RAM 630, an internal storage device 640, an IOC 650, and a NIC 680, and constitutes a computer.
- the CPU 610 reads a program from ROM 620.
- the CPU 610 controls the RAM 630, the internal storage device 640, the IOC 650, and the NIC 680 based on the read program.
- the CPU 610 controls these configurations and implements the functions as the prediction formula set learning generation unit 12000 and the operation amount determination unit 14000 shown in FIG.
- the CPU 610 may use the RAM 630 or the internal storage device 640 as a temporary storage of a program when realizing each function.
- the CPU 610 may read the program included in the storage medium 700 storing the program so as to be readable by a computer using a storage medium reading device (not shown). Alternatively, the CPU 610 may receive a program from an external device (not shown) via the NIC 680.
- ROM 620 stores programs executed by CPU 610 and fixed data.
- the ROM 620 is, for example, a P-ROM (Programmable-ROM) or a flash ROM.
- the RAM 630 temporarily stores programs executed by the CPU 610 and data.
- the RAM 630 is, for example, a D-RAM (Dynamic-RAM).
- the internal storage device 640 stores data and programs stored in the information processing device 600 for a long period of time. Further, the internal storage device 640 may operate as a temporary storage device for the CPU 610.
- the internal storage device 640 is, for example, a hard disk device, a magneto-optical disk device, an SSD (Solid State Drive), or a disk array device.
- the RAM 630 or the internal storage device 640 implements a function as the information storage unit 11000.
- the information processing apparatus 600 may realize the function of the information storage unit 11000 using both the RAM 630 and the internal storage device 640.
- the information storage unit 11000 can be realized by using any one of a volatile memory and a nonvolatile memory.
- the IOC 650 mediates data between the CPU 610, the input device 660, and the display device 670.
- the IOC 650 is, for example, an IO interface card.
- the input device 660 is a device that receives an input instruction from an operator of the information processing apparatus 600.
- the input device 660 is, for example, a keyboard, a mouse, or a touch panel.
- the input device 660 of the information processing apparatus 600 may operate as a part of the input apparatus 22000. That is, the information processing device 600 may include the function of the input device 22000.
- the display device 670 is a device that displays information to the operator of the information processing apparatus 600.
- the display device 670 is a liquid crystal display, for example.
- the display device 670 of the information processing apparatus 600 may operate as a part of the output apparatus 23000. That is, the information processing device 600 may include the function of the output device 23000.
- the NIC 680 relays data exchange with external devices (for example, the input device 22000 and the output device 23000) via the network.
- the NIC 680 is, for example, a LAN (Local Area Network) card.
- the information processing apparatus 600 configured in this way can obtain the same effects as the information processing apparatus 10000.
- FIG. 6 is a block diagram illustrating an example of the configuration of the information processing apparatus 10010 according to the second embodiment.
- the information processing apparatus 10010 of this embodiment includes a prediction formula set storage unit 13000 between the prediction formula set learning generation unit 12000 and the operation amount determination unit 14000 in addition to the configuration of the information processing device 10000 of the first embodiment. .
- Other configurations of the information processing apparatus 10010 are the same as those of the information processing apparatus 10000. Therefore, in this embodiment, a configuration and operation different from those of the first embodiment will be described, and description of the same configuration and operation as those of the first embodiment will be omitted.
- the prediction formula set storage unit 13000 stores the prediction formula set 2500 learned and generated by the prediction formula set learning generation unit 12000.
- the prediction formula set 2500 including the input, output, and stored prediction formula set is referred to as a prediction formula set 2500.
- the operation amount determination unit 14000 of the present embodiment is based on the prediction formula set 2500 stored in the prediction formula set storage unit 13000, the received control target information 2100, the accumulated information 2600, and the input information 2200.
- the operation amount 2300 of the control target 21000 is determined.
- the prediction formula set learning generation unit 12000 and the operation amount determination unit 14000 operate continuously.
- the prediction formula set learning generation unit 12000 and the operation amount determination unit 14000 do not have to operate continuously.
- the prediction equation set learning generation unit 12000 may operate at an interval longer than the operation interval of the operation amount determination unit 14000.
- the operation interval of the operation amount determination unit 14000 may be set to 1 second, whereas the operation interval of the prediction formula set learning generation unit 12000 may be set to 1 day.
- the prediction formula set storage unit 13000 absorbs the difference between the operation interval of the prediction formula set learning generation unit 12000 and the operation interval of the operation amount determination unit 14000 in the above case. Therefore, the prediction formula set storage unit 13000 receives the prediction formula set 2500 learned and generated by the prediction formula set learning generation unit 12000 and stores it.
- the operation amount determination unit 14000 can use the prediction formula set 2500 stored in the prediction formula set storage unit 13000 when the operation amount 2300 is determined. Therefore, the prediction formula set learning generation unit 12000 of this embodiment does not need to learn and generate the prediction formula set 2500 every time the operation amount determination unit 14000 determines the operation amount 2300. That is, the prediction formula set learning generation unit 12000 of the present embodiment can reduce the amount of calculation.
- the information processing apparatus 10010 can obtain an effect that the amount of calculation can be reduced in addition to the effect of the first embodiment.
- the operation amount determination unit 14000 uses the prediction formula set 2500 stored in the prediction formula set storage unit 13000 in determining the operation amount 2300. Therefore, the prediction formula set learning generation unit 12000 of this embodiment does not need to learn and generate the prediction formula set 2500 every time the operation amount determination unit 14000 determines the operation amount 2300. That is, the prediction formula set learning generation unit 12000 of this embodiment can reduce the amount of calculation.
- the information processing apparatus 10010 of this embodiment may correspond to a plurality of control objects 21000 as in the first embodiment.
- the information processing apparatus 10010 of the present embodiment may be realized by a computer shown in FIG. 5 as in the first embodiment.
- FIG. 7 is a block diagram illustrating an example of the configuration of the information processing apparatus 10020 according to the third embodiment.
- the information processing apparatus 10020 of this embodiment includes a predicted value calculation unit 15000 in addition to the configuration of the information processing apparatus 10010 of the second embodiment.
- Other configurations of the information processing apparatus 10020 are the same as those of the information processing apparatus 10010. Therefore, in this embodiment, a configuration and operation different from those of the second embodiment will be described, and description of the same configuration and operation as those of the second embodiment will be omitted. Note that the information processing apparatus 10020 does not need to include the prediction formula set storage unit 13000 as with the information processing apparatus 10000 of the first embodiment.
- the predicted value calculation unit 15000 calculates a predicted value 2400 based on the predicted formula set 2500 stored in the predicted formula set storage unit 13000, the control target information 2100, and the operation amount 2300. For example, the predicted value calculation unit 15000 calculates the predicted value 2400 by applying the control target information 2100 and the operation amount 2300 to the prediction formula set 2500.
- the predicted value calculation unit 15000 may use the accumulated information 2600 to calculate the predicted value 2400 as necessary.
- the predicted value 2400 is a value predicted to shift the control target 21000 when the operation amount 2300 is applied to the control target 21000. That is, the predicted value 2400 is a value corresponding to the result for the operation amount 2300.
- the predicted value calculation unit 15000 may output the calculated predicted value 2400 to the output device 23000, for example. Further, as already described, the operation amount determination unit 14000 may output the operation amount 2300 to the output device 23000. In this case, the operator can confirm both the operation amount 2300 and the predicted value 2400 with reference to the output device 23000.
- the information processing apparatus 10020 can obtain a predicted value 2400 for the operation amount 2300 in addition to the effects of the second embodiment.
- the reason is that the predicted value calculation unit 15000 calculates and outputs the predicted value 2400 based on the prediction formula set 2500, the control target information 2100, and the operation amount 2300.
- the information processing apparatus 10020 of the present embodiment may correspond to a plurality of control objects 21000 as in the first embodiment.
- the predicted value calculation unit 15000 calculates a plurality of predicted values 2400.
- the predicted value calculation unit 15000 may calculate some predicted values 2400 instead of all predicted values 2400.
- the predicted value calculation part 15000 may calculate the predicted value 2400 regarding the control variable designated from the input device 22000, for example.
- the information processing apparatus 10020 of the present embodiment may be realized by a computer shown in FIG. 5 as in the first embodiment.
- FIG. 8 is a block diagram illustrating an example of the configuration of the information processing apparatus 10030 according to the fourth embodiment.
- the information processing apparatus 10030 of this embodiment includes a predicted value accumulation unit 16000, a prediction error calculation unit 17000, and a relearning determination unit 18000 in addition to the configuration of the information processing apparatus 10020 of the third embodiment.
- Other configurations of the information processing apparatus 10030 are the same as those of the information processing apparatus 10020. Therefore, in this embodiment, a configuration and operation different from those of the third embodiment will be described, and description of the same configuration and operation as those of the third embodiment will be omitted. Note that the information processing apparatus 10030 does not need to include the prediction formula set storage unit 13000 as with the information processing apparatus 10000 of the first embodiment.
- the prediction value storage unit 16000 receives the prediction value 2400 calculated by the prediction value calculation unit 15000 and stores it.
- the predicted value 2400 including the input value, the output, and the stored predicted value 2400 is referred to as the predicted value 2400.
- the prediction error calculation unit 17000 calculates a prediction error 2900 based on the storage information 2600 stored in the information storage unit 11000 and the prediction value 2400 stored in the prediction value storage unit 16000.
- the prediction error 2900 is a difference (error) between a value set for the control target 21000 (control target information 2100) and the prediction value 2400.
- the re-learning determination unit 18000 determines whether or not to re-learn in the prediction formula set learning generation unit 12000 based on the prediction error 2900 calculated by the prediction error calculation unit 17000.
- the result of this determination is referred to as “determination result 3000”.
- the relearning determination unit 18000 may refer to other information such as a threshold value in the determination process.
- the prediction formula set learning generation unit 12000 receives the determination result 3000 of the relearning determination unit 18000. Then, when the determination result 3000 indicates relearning, the prediction formula set learning generation unit 12000 learns the prediction formula set 2500 again and regenerates it.
- the relearning determination unit 18000 of the present embodiment outputs a determination result 3000 that relearning is performed when the prediction error 2900 is equal to or greater than a certain value (threshold value), for example.
- a certain value for example.
- the prediction formula set learning generation unit 12000 regenerates the prediction formula set 2500.
- the information processing apparatus 10030 automatically relearns the prediction formula set 2500.
- the information processing apparatus 10030 can maintain high prediction accuracy.
- the information processing apparatus 10030 may execute relearning based on an instruction from the input apparatus 22000 or an external apparatus (not shown). For example, when the surrounding environment 20000 changes significantly and / or when the configuration of the control target 21000 changes significantly, the operator operates the input device 22000 to re-establish the prediction formula set 2500 in the information processing device 10030. Learning and regeneration may be instructed. Based on this instruction, the information processing apparatus 10030 may update the prediction formula set 2500 by causing the prediction formula set learning generation unit 12000 to operate again. As a result, the information processing apparatus 10030 can realize control adapted to changes in the surrounding environment 20000 and / or the control target 21000.
- the information processing apparatus 10030 of the present embodiment can obtain an effect that higher prediction accuracy can be realized in addition to the effect of the third embodiment.
- the prediction error calculation unit 17000 of this embodiment calculates a prediction error 2900 based on the accumulated information 2600 and the prediction value 2400.
- the relearning determination unit 18000 instructs the prediction formula set learning generation unit 12000 to relearn and regenerate the prediction formula set 2500.
- the information processing apparatus 10030 can generate an appropriate prediction formula set 2500.
- the information processing apparatus 10030 of this embodiment may correspond to a plurality of control objects 21000, as in the first embodiment.
- the information processing apparatus 10030 of the present embodiment may be realized by a computer shown in FIG. 5 as in the first embodiment.
- FIG. 9 is a block diagram illustrating an example of the configuration of the operation amount determination unit 14000 according to the present embodiment.
- the operation amount determination unit 14000 includes a prediction formula set conversion unit 14100 and an operation amount calculation unit 14200.
- the prediction formula set conversion unit 14100 receives the prediction formula set 2500. Then, the prediction formula set conversion unit 14100 converts the prediction formula set 2500 into the prediction formula set 2501 according to a predetermined rule so that the operation amount 2300 can be easily calculated by the operation amount calculation unit 14200 described later. For example, the prediction formula set conversion unit 14100 may use a conversion process that simplifies the prediction formula set 2500. That is, the prediction formula set conversion unit 14100 deletes a variable related to the operation amount 2300 having a small change in the predicted value 2400 with respect to the change in the operation amount 2300 from the prediction formula set 2500, thereby simplifying the prediction formula set. You may convert into 2501. Specifically, for example, the prediction formula set conversion unit 14100 may delete a variable having a small coefficient in the prediction formula.
- the prediction formula set conversion unit 14100 may convert the prediction formula set 2500 into the prediction formula set 2501 using the priority included in the input information 2200 input from the input device 22000. Specifically, for example, the prediction formula set conversion unit 14100 does not simplify the prediction formula set 2500 that predicts a control variable with a high priority, but simplifies the prediction formula set 2500 that predicts a control variable with a low priority. May be. Alternatively, the prediction formula set conversion unit 14100 may simplify many control variables with low priority and simplify the control variables with high priority.
- the operation amount calculation unit 14200 constructs a prediction control model based on the prediction formula set 2501 converted by the prediction formula set conversion unit 14100, the control target information 2100, the accumulated information 2600, and the input information 2200, and operates The quantity 2300 is calculated.
- FIG. 10 is a block diagram illustrating an example of the configuration of the operation amount calculation unit 14200 according to the present embodiment.
- the manipulated variable calculation unit 14200 includes a mathematical programming problem formulation unit 14210 and a mathematical programming problem calculation unit 14220.
- the mathematical programming problem formulation unit 14210 constructs a prediction control model based on the prediction formula set 2501 converted by the prediction formula set conversion unit 14100, the control target information 2100, the accumulated information 2600, and the input information 2200.
- the mathematical programming problem 2800 is established (formulated).
- the mathematical programming problem 2800 is a mathematical expression (description) of the predictive control model, and is, for example, a linear programming problem, a quadratic programming problem, a combinatorial optimization problem, a linear integer problem, or a mixed integer programming problem.
- the mathematical programming problem calculation unit 14220 calculates (solves) the mathematical programming problem 2800 formulated by the mathematical programming problem formulation unit 14210 using an appropriate solver, and calculates the manipulated variable 2300.
- the solver is a configuration for calculating an optimal solution of the formulated mathematical programming problem 2800.
- the solver may be, for example, a dedicated computer or a program executed on a computer.
- FIG. 11 is a block diagram illustrating an example of another configuration of the operation amount calculation unit 14200.
- the operation amount calculation unit 14201 shown in FIG. 11 includes a mathematical programming problem formulation unit 14210, a mathematical programming problem calculation unit 14220, an operation amount good / bad determination unit 14230, and an operation amount selection unit 14240.
- the mathematical programming problem formulation unit 14210 and the mathematical programming problem calculation unit 14220 are the same as the mathematical programming problem formulation unit 14210 and the mathematical programming problem calculation unit 14220 shown in FIG. However, the mathematical programming problem calculation unit 14220 may output a plurality of calculation results 2700.
- the predicted value calculation unit 15000 calculates a predicted value 2400 based on the calculation result 2700 of the mathematical programming problem calculation unit 14220.
- the prediction value calculation unit 15000 calculates a prediction value 2400 for each calculation result 2700. Then, for example, the calculation result 2700 of the mathematical programming problem calculation unit 14220 and the prediction value 2400 calculated by the prediction value calculation unit 15000 are output to the output device 23000.
- the operation amount pass / fail determination unit 14230 calculates the calculation result 2700 (this is the operation amount 2300). Is determined).
- the output device 23000 displays the calculation result 2700 and the predicted value 2400. Then, the operator refers to the value displayed on the output device 23000 and determines whether the calculation result 2700 (corresponding to the operation amount 2300) and the predicted value 2400 are appropriate. Then, using the input device 22000, the operator transmits “good / bad determination” to the operation amount calculation unit 14201.
- the operation amount pass / fail determination unit 14230 may hold a reference value of “pass / fail determination” (for example, a threshold value or a range that is good) and determine pass / fail based on the reference value.
- a reference value of “pass / fail determination” for example, a threshold value or a range that is good
- the operation amount good / bad determination unit 14230 instructs the mathematical programming problem formulation unit 14210 to re-formulate the mathematical programming problem again.
- the mathematical programming problem formulation unit 14210 receives new input information 2200 (for example, a set value, a constraint, or a priority) so that an appropriate operation amount 2300 can be calculated. . Therefore, the operation amount calculation unit 14201 may request the operator to input again via the output device 23000.
- the operation amount good / bad determination unit 14230 sends the calculation result 2700 to the operation amount selection unit 14240.
- the operation amount selection unit 14240 selects an appropriate operation amount 2300.
- the operation amount selection unit 14240 selects the calculation result 2700 as the operation amount 2300.
- the operation amount selection unit 14240 selects the calculation result 2700 as the operation amount 2300 based on the input information 2200 from the input device 22000.
- the operation amount selection unit 14240 may select an appropriate calculation result 2700 as the operation amount 2300 according to a predetermined rule.
- the predetermined rule is, for example, a rule that the difference from the previous operation amount 2300 is the smallest or a rule that the predicted cost is the smallest.
- FIG. 12 is a block diagram showing an example of the configuration of the mathematical programming problem formulation unit 14210 according to this embodiment.
- the mathematical programming problem formulation unit 14210 shown in FIG. 12 includes a constraint equation generation unit 14212, an objective function generation unit 14214, an allowable error register 14216, an error weight coefficient register 14217, and a cost coefficient register 14218.
- the constraint expression generation unit 14212 is based on the prediction expression set 2501 converted by the prediction expression set conversion unit 14100, the control target information 2100, the accumulated information 2600, the input information 2200, or the allowable error described later. , Generate constraint expressions. However, the input information 2200 includes at least one of a setting value or a constraint.
- the allowable error register 14216 holds an allowable error that is a maximum difference recognized between the set value and the predicted value 2400. For example, when a predicted value 2400 from 8 to 12 with respect to the setting value of 10 is recognized, the allowable error is “ ⁇ 2”.
- the error weight coefficient register 14217 holds an error weight coefficient that is a coefficient for handling the priority provided for the plurality of control objects 21000 as an error weight. For example, it is assumed that there are two control objects A and B. Then, it is assumed that the control of the control target A has higher priority than the control of the control target B. In this case, the error weight coefficient register 14217 holds an error weight coefficient larger than the error weight coefficient for the control target B as the error weight coefficient for the control target A. Based on such an operation, the mathematical programming problem formulation unit 14210 enables formulation in consideration of priority. For example, the mathematical programming problem formulation unit 14210 can give a larger difference in priority as the difference in error weighting coefficient is larger.
- the mathematical programming problem formulation unit 14210 may not receive the priority from the input device 22000. In that case, the error weight coefficient register 14217 may hold all error weight coefficients as the same value other than 0. Alternatively, in that case, the mathematical programming problem formulation unit 14210 may not use the error weight coefficient register 14217.
- the cost coefficient register 14218 holds a cost coefficient that is a coefficient for uniformly handling a plurality of cost indexes.
- the amount is a value that can be handled in a unified manner. Therefore, when the electricity usage amount and the gas usage amount are predicted as costs, the cost coefficient register 14218 converts each usage amount into a monetary amount by multiplying the usage amount by the unit price, and holds it. Then, the mathematical programming problem formulation unit 14210 can handle the cost indexes of the electric usage amount and the gas usage amount in a unified manner.
- the objective function generation unit 14214 is based on the prediction formula set 2501 converted by the prediction formula set conversion unit 14100, the control target information 2100, the accumulated information 2600, the allowable error, the error weighting coefficient, and the cost coefficient. Generate an objective function.
- the mathematical programming problem formulation unit 14210 may receive the allowable error and the cost coefficient from the input device 22000.
- the first effect is that the amount of calculation can be reduced.
- the reason is that the prediction formula set conversion unit 14100 simplifies the prediction formula set 2500 used for calculation.
- the second effect can be predicted more accurately.
- the reason is that the operation amount pass / fail judgment unit 14230 can instruct re-operation of the mathematical programming problem formulation unit 14210 based on the “pass / fail judgment”.
- FIG. 14 is a block diagram illustrating an example of the configuration of the information processing apparatus 10040 according to the sixth embodiment.
- the information processing apparatus 10040 of this embodiment includes an information storage unit 11000, a prediction control unit 30000, a fixed control unit 31000, and an operation amount selection unit 32000.
- the information storage unit 11000 is the same as in the first to fifth embodiments.
- the prediction control unit 30000 includes the prediction formula set learning generation unit 12000 and the operation amount determination unit 14000 of the first to fifth embodiments, and realizes the operation of the first to fifth embodiments.
- Fixed control unit 31000 calculates an operation amount according to a preset control equation.
- the operation amount selection unit 32000 selects and outputs one control amount from a plurality of control amounts.
- the information storage unit 11000 and the prediction control unit 30000 perform the operations described in the first to fifth embodiments, and output an operation amount based on the prediction control.
- This operation amount is the first operation amount.
- control target information 2100 is not directly input to the operation amount determination unit 14000 but is used as the storage information 2600 via the information storage unit 11000.
- this is an example of the information flow.
- the control target information 2100 may be directly input.
- the control target information 2100 may be input via the information storage unit 11000.
- the fixed control unit 31000 calculates based on the control target information 2100 and the accumulated information 2600 based on the control formula input in advance by the designer, and outputs the operation amount. This operation amount is the second operation amount.
- the fixed control unit 31000 configures a control expression by a relational expression of an input (information on surrounding environment and measurement target), a set value, and an output (operation amount).
- the fixed control unit 31000 can take a formula of a control method such as feedback or feedforward obtained by proportional control, classical control of PID (Proportional Integral Derivative) control, or modern control obtained from a state equation. That's fine.
- the prediction control unit 30000 and the fixed control unit 31000 are not limited to one and may include a plurality.
- the prediction control unit 30000 may have different characteristics based on the randomness of the initial value. Therefore, based on the provision of a plurality of prediction control units 30000, the information processing apparatus 10040 can output an operation amount having different characteristics. Further, a plurality of fixed control units 31000 may be included when the designer prepares different control expressions in terms of followability or stability.
- the operation amount selection unit 32000 selects which of the operation amount determination unit 14000 and the fixed control unit 31000 to use.
- the information processing apparatus 10040 may select one of the control target information 2100, the accumulated information 2600, the operation amount output from the operation amount determination unit 14000, Alternatively, the display unit 33000 may be included. Then, the operation amount selection unit 32000 selects an operation amount of the prediction control unit 30000 or the fixed control unit 31000 based on an input (selection information) from the user, and outputs the operation amount.
- the information processing apparatus 10040 may include a prediction error determination unit 34000 that uses the prediction value calculation unit 15000 described in the third embodiment.
- the prediction error determination unit 34000 calculates a prediction value using the prediction value calculation unit 15000 and compares it with the information of the accumulated information 2600 to obtain a prediction error. Then, the operation amount selection unit 32000 determines which operation amount of the prediction control unit 30000 or the fixed control unit 31000 to use based on the magnitude of the error.
- the prediction error determination unit 34000 selects, for example, the one with the smallest error from the prediction control unit 30000. Then, the operation amount selection unit 32000 selects the operation amount of the prediction control unit 30000 having the smallest error when the error is equal to or less than a certain value.
- the operation amount selection unit 32000 selects the fixed control unit 31000.
- the operation amount selection unit 32000 outputs the selection result. Based on this operation, when the prediction accuracy is sufficient, that is, when the accuracy of the prediction control unit 30000 is high, the prediction control unit 30000 is selected. Therefore, the information processing apparatus 10040 can perform highly accurate control.
- the information processing apparatus 10040 evaluates the operation amount based on the operation amount of the operation amount determination unit 14000 or the fixed control unit 31000 and selects which operation amount to use.
- the operation amount evaluation unit 35000 may be included.
- the operation amount selection unit 32000 of the information processing apparatus 10040 may select the operation amount based on the selection result of the operation amount evaluation unit 35000.
- the operation amount evaluation unit 35000 inputs the operation amount into an evaluation function input in advance to the operation amount evaluation unit 35000.
- the operation amount evaluation unit 35000 outputs the operation amount that provides the best result as the selection result.
- the evaluation function may be, for example, a function for obtaining the cost for the operation amount. In this case, the information processing apparatus 10040 can select the control with the lowest cost and can control the control target at a low cost.
- the information processing apparatus 10040 evaluates the operation to be controlled based on the information of the accumulated information 2600 and the input information 2200 and outputs a selection result. May be included.
- the operation amount selection unit 32000 of the information processing apparatus 10040 selects an operation amount based on the selection result of the motion evaluation unit 36000.
- the operation evaluation unit 36000 compares the set value of the control target 21000 of the input information with the accumulated information 2600. Then, the motion evaluation unit 36000 evaluates whether or not the control target is operating within the set error range according to the set value. At this time, when the control target is moving according to the set value, the operation amount selection unit 32000 selects the operation amount of the prediction control unit 30000.
- the operation amount selection unit 32000 selects the fixed control unit 31000. Based on this operation, when the control characteristics of the prediction control unit 30000 are not sufficient, the information processing apparatus 10040 can perform control based on the fixed control unit 31000, and stability can be increased.
- the information processing apparatus 10040 includes a prediction formula set learning generation unit 12000, and evaluates the prediction formula based on the learning evaluation value of the prediction formula set learning generation unit 12000. You may select a result from the selection result of the prediction formula evaluation part 37000 to do.
- the prediction formula evaluation unit 37000 operates as follows.
- the prediction formula set learning generation unit 12000 has an evaluation value inside for learning.
- An example of the evaluation value is an evaluation value based on an information amount.
- the operation amount selection unit 32000 determines that the prediction formula is sufficient and selects the prediction control unit 30000 to be used. Further, the operation amount selection unit 32000 selects the fixed control unit 31000 in cases other than the above.
- this embodiment can obtain the effect of realizing more accurate control.
- the reason is that the operation amount selection unit 32000 of the information processing apparatus 10040 appropriately selects the first operation amount that is the output of the prediction control unit 30000 and the second operation amount that is the output of the fixed control unit 31000. Because.
- the prediction formula set 2500 learned and generated by the prediction formula set learning generation unit 12000 is a set of discrete-time linear functions.
- control variables of the control target 21000 is “n y ”, and the control variables are expressed as “y 1 , y 2 ,.
- state variables representing the states of the peripheral environment 20000 and the controlled object 21000
- n x The number of variables representing the states of the peripheral environment 20000 and the controlled object 21000
- state variables are expressed as “x 1 , x 2 ,.
- x 1 , x 2 there is a weather forecast as an example of a state variable.
- the weather forecast obtained at the present time in the weather forecast is interpreted as a current state variable and not as a future state variable.
- the state variable corresponds to the control target information 2100.
- cost variables the number of variables related to the cost of the controlled object 21000 (hereinafter referred to as “cost variables”) among the state variables is “n c ”, and the cost variables are “c 1 , c 2 ,. Express.
- operation variables representing the operation amount to be added to the control target 21000 is “n u ”, and the operation variables are expressed as “u 1 , u 2 ,.
- Formula 1 has the following format. That is, first, Equation 1 multiplies the state variables from the past to the present from n steps before to k steps by the coefficient for each variable. Then, Equation 1 multiplies the operation variable from the past from the nth step to the first step before the kth step to immediately before the present by the coefficient for each variable. Then, Equation 1 multiplies the operation variables from the present to the future from the k-th step to one step before the i-step by a coefficient for each variable. Formula 1 is a form in which a constant is added to the result.
- Equation 1 can be rewritten as Equation 2 using a matrix. [Equation 2]
- T indicates a transposed matrix
- Equation 3 can be simplified and described by Equation 4 below. [Equation 4]
- the prediction formula set learning generation unit 12000 converts the prediction formula set 2500 having the cost variables “c 1 , c 2 ,...” After i steps at the k-th step as dependent variables into the matrix A x, c, i. , A u, c, i and B c, i can be generated by the following Equation 5. [Equation 5]
- Equation 4 and Equation 5 do not include the following variables.
- a non-zero element in a column of a matrix [A u, y, i A u, c, i ] T configured by connecting the matrix A u, y, i and the matrix A u, c, i Is 2 or more.
- At least one of the elements is an element of a column of the matrix A u, y, i .
- a non-zero element being an element of a column of the matrix A u, y, i means that an operation variable having the element as a coefficient can change a control variable.
- the presence of a plurality of such non-zero elements means that a plurality of control variables can be changed simultaneously based on a single manipulated variable. That is, the relationship between the control variables can be described.
- the non-zero element is an element of the column of the matrix A u, c, i means that the manipulated variable also affects the cost variable. That is, the relationship between the control variable and the cost variable can be described.
- the prediction formula set 2500 includes Formula 4 and Formula 5.
- the prediction formula set conversion unit 14100 (FIG. 9) may simplify the prediction formula set 2500 (create a prediction formula set 2501) as appropriate.
- the mathematical programming problem formulation unit 14210 (FIGS. 10 to 12) can construct a predictive control model using the predictive formula set 2501.
- Equation 6 information (set value (Y SV (k + i))) input from the input device 22000 for the future control variables “y 1 , y 2 ,... This is described by Equation 6. However, in the present invention, it is not necessary to give the corresponding set values (y 1, SV (k + i), y 2, SV (k + i), etc To all control variables, and some set values are given. Also good. [Equation 6]
- Equation 7 When a set value is given to the control variable “y 1 , y 2 ,...” After the i-th step in the k-th step, the allowable error (E (k + i)) is described by the following Equation 7. The However, according to the present invention, it is not necessary to provide tolerances (e 1 (k + i), e 2 (k + i), etc Corresponding to all control variables, and some tolerances may be given. [Equation 7]
- Equation 8 The error weighting coefficient (W e (k + i)) with respect to the allowable error described by Equation 7 is described by Equation 8 below.
- the error weighting coefficient is determined based on the input information 2200 (priority) from the input device 22000.
- Corresponding to the control variable is set large for the control variable having a high priority.
- the mathematical programming problem formulation unit 14210 can determine an operation variable in consideration of the priority. [Equation 8]
- the cost coefficient (W c (k + i)) is described by the following mathematical formula 9.
- the cost coefficient is a coefficient (w C1 (k + i), w C2 (k + i),...)
- w C1 (k + i) For handling a plurality of cost variables “c 1 , c 2 ,.
- the input information 2200 (constraint) input from the input device 22000 with respect to the control variable “y 1 , y 2 ,.
- the minimum-side constraint is “Y MIN (k + i)”
- the maximum-side constraint is “Y MAX (k + i)”.
- Constraints corresponding to control variables ("y1 , MIN (k + i), y2 , MIN (k + i), ## and "y1 , MAX (k + i), y2 , MAX (k + i), ##) are necessary.
- minus infinity may be set as the lower limit, or plus infinity may be set as the upper limit. [Equation 10]
- Equation 11 the input information 2200 (constraint) input from the input device 22000 for the operation variables “u 1 , u 2 ,... Described.
- the minimum-side constraint is “U MIN (k, k + i ⁇ 1)”
- the maximum-side constraint is “U MAX (k, k + i ⁇ 1)”.
- constraints corresponding to the control variables (“u 1, MIN (k), u 1, MIN (k + 1), etc.
- And“ u 1, MAX (k), u 1, MAX (k + 1),. )) May have a lower limit of minus infinity or an upper limit of plus infinity.
- the magnitude relationship represented using the matrix in Formula 12 indicates the magnitude relationship between each element of the matrix.
- the first element of the first magnitude relation equation of Equation 12 is “y 1, SV (k + i) ⁇ e 1 (k + i) ⁇ y 1 (k + i) ⁇ y 1, SV (k + i) + e 1 (k + i). ) ”.
- Mathematical programming problem calculation unit 14220 may obtain a vector U (k, k + i ⁇ 1) of operation variables that minimizes the objective function of Expression 13 using Expression 12 as a constraint expression.
- Equation 6 the setting value after i step at the k step is given.
- the set values from the 1st step of the k-th step to the i-th step may be given continuously or intermittently.
- the mathematical programming problem formulation unit 14210 uses the prediction formula set 2500 learned and generated by the prediction formula set learning generation unit 12000 to construct a prediction control model for control.
- the information processing apparatus 10000 can perform predictive control corresponding to the input information 2200 (for example, set values, constraints, and priorities).
- FIG. 13 shows specific values of these equations.
- the matrix [A u, y, 1 A u, c, 1 ] T [1 ⁇ 2 1] T column formed by connecting the matrix Au, y, 1 and the matrix A u, c, 1
- the elements of are all non-zero elements.
- the operation variable u 1 (k) can change all of the two control variables “y 1 (k + 1), y 2 (k + 1)” and one cost variable c 1 (k + 1).
- These formulas are examples of the prediction formula set 2500 in which the operation variable can be determined in consideration of the cost.
- the prediction formula and objective function described above are merely examples.
- the present invention is not limited to these.
- the prediction formula may be a nonlinear function.
- the objective function may be a nonlinear function.
- the mathematical programming problem calculation unit 14220 of the present invention may solve the problem by using the solving method and solver of the mathematical programming problem 2800 that matches the properties of the formulated mathematical programming problem 2800.
- Information storage means for receiving and storing control target information including information on the control target and the surrounding environment including the control target; and Prediction formula set learning generating means for learning and generating a prediction formula set used for determining an operation amount for the control target based on the control target information stored in the information storage means;
- the input information necessary for determining the operation amount of the control target is received, based on the prediction formula set, the control target information stored in the information storage unit, the received control target information, and the input information,
- An operation amount determination means for constructing a predictive control model of the control object and determining an operation amount used for the control of the control object; Including information processing apparatus.
- the prediction formula set is Including two or more prediction formulas having a state variable representing a state of a future predictive control model or a control variable to be controlled by a future predictive control model as a dependent variable,
- the information processing apparatus according to claim 1, further comprising: a second prediction formula that uses an operation variable representing an operation of at least one prediction control model after the present as a common independent variable with the first prediction formula.
- the manipulated variable determining means is The information processing apparatus according to claim 1 or 2, wherein as the input information, one or both of a set value and a constraint for a control variable to be controlled in the predictive control model is received.
- a prediction formula set storage unit that stores the prediction formula set generated by the prediction set learning generation unit;
- the manipulated variable determining means is The information processing apparatus according to any one of Supplementary Note 1 to Supplementary Note 4, wherein a prediction formula set stored in the prediction formula set storage unit is used to determine an operation amount to the control target.
- Appendix 6 Further including prediction value calculation means for calculating a prediction value of the control object based on the prediction expression set stored in the prediction expression set storage means, the received control target information, and the manipulated variable The information processing apparatus according to any one of appendix 5.
- the predicted value calculation means The information processing apparatus according to appendix 6, wherein the control target information stored in the information storage unit is used for calculating the predicted value.
- Predicted value storage means for storing the predicted value
- a prediction error calculation means for calculating a prediction error based on the prediction value stored in the prediction value storage means and the control target information stored in the information storage means
- Re-learning determination means for determining whether to re-learn in the prediction formula set learning generation means using the prediction error
- the prediction formula set learning generation means includes: The information processing apparatus according to appendix 6 or appendix 7, wherein the prediction formula set is relearned and regenerated based on a determination result of the relearning determination unit.
- the information storage means receives and stores a plurality of pieces of control target information;
- the information processing apparatus according to any one of attachments 1 to 8, wherein the operation amount determination unit determines operation amounts of the plurality of control targets.
- the manipulated variable determining means is Prediction formula set conversion means for converting the prediction formula set according to a predetermined rule; Based on the prediction formula set converted by the prediction formula set conversion unit, the control target information stored in the information storage unit, the received control target information, and the input information, the operation amount to the control target is calculated.
- the information processing apparatus according to any one of appendices 1 to 9, further comprising:
- the operation amount calculation means is A mathematical programming problem representing the predictive control model based on the predictive formula set converted by the predictive formula set converting means, the control target information stored in the information storing means, the received control target information, and the input information.
- Mathematical programming problem formulation means to formulate The information processing apparatus according to claim 10, further comprising: a mathematical programming problem calculation unit that calculates the mathematical programming problem formulated and determines the calculation result as the operation amount.
- Appendix 12 An operation amount pass / fail judgment means for judging pass / fail of the calculation result of the mathematical programming problem calculation means;
- the information processing apparatus according to appendix 11, further comprising: an operation amount selection unit that determines the calculation result as the operation amount when the determination result of the operation amount determination unit is good.
- the operation amount selection means is When the operation amount determination means determines that a plurality of calculation results are good, The information processing apparatus according to claim 12, wherein the operation amount is selected from the calculation result based on the input information or a predetermined rule.
- Appendix 14 The information processing apparatus according to appendix 12 or 13, wherein when the manipulated variable quality determination unit determines that all of the calculation results are NO, the mathematical programming problem formulation unit instructs the formulation again.
- the prediction formula set learning generation means includes: As the prediction formula set, Including multiple functions with future state variables or future control variables as dependent variables, One function is that the dependent variable is a future control variable, the independent variable does not include the future state variable and the future control variable,
- the information processing apparatus according to any one of Supplementary Note 1 to Supplementary Note 15, including another function that uses the one function and at least one operation variable after the present as a common independent variable.
- Control target information including information on the control target and the surrounding environment including the control target, storing the control target information; Learning and generating a set of prediction formulas used to determine the operation amount of the control target based on the accumulated control target information, Receiving input information necessary for determining the manipulated variable of the control target, and predicting the control target based on the prediction formula set, the accumulated control target information, the received control target information, and the input information
- Appendix 18 A process of receiving control target information including information on a control target and surrounding environment including the control target, and storing the control target information; Learning and generating a prediction formula set used for determining the operation amount of the control target based on the accumulated control target information; Receiving input information necessary for determining the manipulated variable of the control target, and predicting the control target based on the prediction formula set, the accumulated control target information, the received control target information, and the input information
- the computer-readable recording medium which recorded the program which builds a control model and makes a computer perform the process which determines the operation amount which uses the said control object for control.
- Information storage means for receiving and storing control target information including information on the control target and the surrounding environment including the control target; and Necessary for determining the manipulated variable of the control object, and the predictive expression set learning generating means for learning and generating the predictive expression set used for determining the manipulated variable for the controlled object based on the control object information accumulated in the information accumulating means
- Input control information constructing a predictive control model of the control target based on the input information, the control target information stored in the information storage means, and the prediction formula set, to control the control target
- An operation amount determining means for determining an operation amount to be used;
- a fixed control means for determining an operation amount used for controlling the control object based on a predictive control means including a formula inputted in advance and control target information stored in the information storage means, and the operation amount determination means
- An operation amount selecting unit that selects any one of the operation amount and the operation amount of the fixed control unit.
- the operation amount selection means is The information processing apparatus according to claim 19, wherein one of the operation amounts output from the one or more prediction control means and the fixed control means is selected.
- Appendix 21 Display means for displaying control target information stored by the information storage means, and operation amounts output from the operation amount determination means and the fixed control means, to the user;
- the information processing apparatus according to appendix 19 or appendix 20, wherein the operation amount selection unit outputs an operation amount selected by a user.
- (Appendix 22) Means for calculating a prediction value based on a prediction expression set output by the prediction expression set learning generation means and control target information stored by the information storage means; Prediction error determination means for determining a prediction error based on a difference between the predicted value and control target information stored in the information storage means; The information according to appendix 19 or appendix 20, wherein the operation amount selection unit selects one of the operation amounts output from the prediction control unit and the fixed control unit based on the result of the prediction error determination unit. Processing equipment.
- the evaluation value is calculated based on a previously input evaluation function, and includes a control evaluation unit that determines the control amount with the highest evaluation value,
- control evaluation unit determines based on an amount of deviation of an operation amount of the operation amount determination unit from a control amount of the fixed control unit.
- An operation evaluation unit that evaluates an operation based on a difference between control target information stored in the information storage unit and a target value of input information, and the operation amount selection unit is configured to perform the control evaluation based on the evaluation of the operation evaluation unit.
- the information processing apparatus according to appendix 19 or appendix 20, wherein one of the operation amounts output from the prediction control unit and the fixed control unit is selected based on a result of the unit.
- Appendix 27 Based on the learning evaluation value calculated by the prediction formula set learning generation unit, including a prediction formula evaluation means for performing an evaluation, The information according to appendix 19 or appendix 20, wherein the operation amount selection unit selects one of the operation amounts output from the prediction control unit and the fixed control unit based on a result of the prediction formula evaluation unit. Processing equipment.
- Appendix 28 28.
- a prediction expression set evaluation unit that evaluates the prediction expression set based on whether or not the prediction expression set output from the prediction expression set learning generation unit is a controllable prediction expression;
- the information processing apparatus according to appendix 19 or appendix 20, wherein one of the operation amounts output from the prediction control unit and the fixed control unit is selected based on a result of the expression set evaluation unit.
- (Appendix 31) Receiving and accumulating control object information including information on the control object and the surrounding environment including the control object; Learning and generating a set of prediction formulas used to determine an operation amount for the control object based on the accumulated control object information, Receiving input information necessary for determining the manipulated variable of the control target, based on the input information, the accumulated control target information and the prediction formula set, construct a predictive control model of the control target, Determining a first operation amount used for controlling the control object; Outputting the first operation amount; Based on the formula inputted in advance and the accumulated control target information, the second operation amount used for the control of the control target is determined and output, A prediction control method that selects one of the first operation amount and the second operation amount.
- (Appendix 32) Processing for receiving and storing control target information including information on the control target and the surrounding environment including the control target; and A process of learning and generating a prediction formula set used for determining an operation amount for the control target based on the accumulated control target information; Receiving input information necessary for determining the manipulated variable of the control target, based on the input information, the accumulated control target information and the prediction formula set, construct a predictive control model of the control target, A process of determining a first operation amount used for controlling the control object; Processing for outputting the first manipulated variable; A process of determining and outputting a second operation amount used for control of the control object based on a formula inputted in advance and the accumulated control object information; A computer-readable recording medium storing a program for causing a computer to execute a process of selecting one of the first operation amount and the second operation amount.
- the information processing apparatus of the present invention can be applied not only to industrial process control, but also to environmental control of buildings and living spaces, and environmental control of agriculture, forestry and fisheries, livestock industry, and the like.
- the information processing apparatus of the present invention can also be applied to actuation control of manipulators, land mobile bodies, water / underwater mobile bodies, air mobile bodies, and the like.
- the information processing apparatus of the present invention can be applied to uses such as communication, traffic, ordering, and financial transactions.
- Information processing system 101 Information processing system 102
- Information processing system 600 Information processing apparatus 610 CPU 620 ROM 630 RAM 640 Internal storage device 650 IOC 660 Input device 670 Display device 680 NIC 700 Storage medium 2100 Control target information 2200 Input information 2300 Manipulated amount 2400 Predicted value 2500 Prediction formula set 2501 Prediction formula set 2600 Accumulated information 2700 Calculation result 2800 Mathematical programming problem 2900 Prediction error 3000 Determination result 10000 Information processing device 10010 Information processing device 10020 Information Processing device 10030 Information processing device 11000 Information accumulation unit 12000 Prediction formula set learning generation unit 13000 Prediction formula set storage unit 14000 Operation amount determination unit 14100 Prediction formula set conversion unit 14200 Operation amount calculation unit 14201 Operation amount calculation unit 14210 Mathematical programming problem formulation Unit 14212 Constraint expression generation unit 14214 Objective function generation unit 14216 Allowable error register 14217 Error weight coefficient register 14218 Cost coefficient register 14220 Number Planning problem calculation unit 14230 Opera
Landscapes
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
Description
図1は、本発明における第1の実施形態の情報処理装置10000を含む情報処理システム100の構成の一例を示すブロック図である。
本実施形態の情報処理装置10000は、制御対象21000の数に制限はない。
情報処理システム100は、情報処理装置10000が決定した操作量2300を、制御対象21000に送信しても良い。
情報処理装置10000構成は、これまでの説明に限らない。
図6は、第2の実施形態に係る情報処理装置10010の構成の一例を示すブロック図である。
図7は、第3の実施形態に係る情報処理装置10020の構成の一例を示すブロック図である。
図8は、第4の実施形態に係る情報処理装置10030の構成の一例を示すブロック図である。
次に、操作量決定部14000のより詳細な実施形態について、第5の実施形態として、図面を参照して説明する。
図14は、第6の実施形態に係る情報処理装置10040の構成の一例を示すブロック図である。
次に、第1乃至第5の実施形態における予測式集合学習生成部12000で学習及び生成する予測式集合2500を生成するための動作例と、操作量決定部14000の動作例について説明する。ここで、操作量決定部140000の動作例は、図9乃至図12に示す第5の実施形態における数理計画問題定式化部14210における定式化の例を、数式を用いて説明する。
すなわち、数式1は、次の形式である。すなわち、まず、数式1は、nステップ前からkステップ目までの過去から現在までの状態変数に変数ごとの係数を乗じる。そして、数式1は、nステップ前からkステップ目の1ステップ前までの過去から現在の直前までの操作変数に変数ごとの係数を乗じる。そして、数式1は、kステップ目からiステップの1ステップ前までの現在から未来までの操作変数に変数ごとの係数を乗じる。そして、数式1は、その結果に定数を加えた形式である。
[数3]
[数5]
(2)未来のコスト変数「cj(k+l) (j=1、…、nc、l=1、…、i)」
(3)未来の制御変数「yj(k+l) (j=1、…、ny、l=1、…、i)」
つまり、ある未来の制御変数と未来のコスト変数とは、他の未来の状態変数、未来のコスト変数、及び、未来の制御変数に依存しない。このことは、後述の制約(数式12)を簡潔に記述できる効果がある。その結果、数理計画問題2800の解法処理が、容易になる。
[数6]
[数7]
[数8]
[数9]
[数10]
[数11]
制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積する情報蓄積手段と、
前記情報蓄積手段に蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成する予測式集合学習生成手段と、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記予測式集合と前記情報蓄積手段に蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる操作量を決定する操作量決定手段と、
含む情報処理装置。
前記予測式集合が、
未来の予測制御モデルの状態を表す状態変数又は未来の予測制御モデルの制御対象となる制御変数を従属変数とする二つ以上の予測式を含み、
前記予測式として、
独立変数が現在以降の予測制御モデルの操作を表す操作変数を含み、未来の状態変数及び未来の制御変数を含まず、従属変数が未来の制御変数である第一の予測式と、
少なくとも一つの現在以降の予測制御モデルの操作を表す操作変数を前記第一の予測式と共通の独立変数とする第二の予測式と
を含む付記1に記載の情報処理装置。
前記操作量決定手段が、
前記入力情報として、前記予測制御モデルにおいて制御の対象となる制御変数に対する設定値又は制約のいずれか又は両方を受信する
付記1又は2に記載の情報処理装置。
前記操作量決定手段が
前記入力情報として、
さらに、複数の前記制御変数における重要性を表す優先度を受信する
付記3に記載の情報処理装置。
前記予測集合学習生成手段が生成した予測式集合を格納する予測式集合格納手段をさらに含み、
前記操作量決定手段が、
前記制御対象への操作量を決定に、前記予測式集合格納手段に格納された予測式集合を用いる
付記1乃至付記4のいずれか1つに記載の情報処理装置。
前記予測式集合格納手段に格納された予測式集合と、前記受信した制御対象情報と、前記操作量とを基に、前記制御対象の予測値を算出する予測値算出手段をさらに含む
付記1乃至付記5のいずれか1つに記載の情報処理装置。
前記予測値算出手段が、
前記予測値の算出に前記情報蓄積手段に蓄積された制御対象情報を用いる
付記6に記載の情報処理装置。
前記予測値を蓄積する予測値蓄積手段と、
前記予測値蓄積手段に蓄積された予測値と前記情報蓄積手段に蓄積された制御対象情報とを基に予測誤差を算出する予測誤差算出手段と、
前記予測誤差を用いて前記予測式集合学習生成手段において再学習するか否か判定する再学習判定手段と
をさらに含み、
前記予測式集合学習生成手段が、
前記再学習判定手段の判定結果に基づいて前記予測式集合を再学習及び再生成する
付記6又は付記7に記載の情報処理装置。
前記情報蓄積手段が、複数の制御対象情報を受信し、蓄積し、
前記操作量決定手段が、前記複数の制御対象の操作量を決定する
付記1乃至8のいずれか1つに記載の情報処理装置。
前記操作量決定手段が、
前記予測式集合を所定の規則に従い変換する予測式集合変換手段と、
前記予測式集合変換手段で変換された予測式集合と前記情報蓄積手段に蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象への操作量を算出する操作量算出手段と
を含む付記1乃至9のいずれか1つに記載の情報処理装置。
前記操作量算出手段が、
前記予測式集合変換手段で変換された予測式集合と前記情報蓄積手段に蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に前記予測制御モデルを表す数理計画問題を定式化する数理計画問題定式化手段と、
定式化された前記数理計画問題を計算し、計算結果を前記操作量と決定する数理計画問題計算手段と
を含む付記10に記載の情報処理装置。
前記数理計画問題計算手段の計算結果の良否を判定する操作量良否判定手段と、
前記操作量良否判定手段の判定結果が良の場合に前記計算結果を前記操作量と決定する操作量選択手段とを含む付記11に記載の情報処理装置。
前記操作量選択手段が、
前記操作量判定手段が複数の計算結果を良と判定した場合に、
前記入力情報又は所定の規則を基に前記計算結果から前記操作量を選択する
付記12に記載の情報処理装置。
前記操作量良否判定手段が
前記計算結果をすべて否と判定した場合に、前記数理計画問題定式化手段に再度の定式化を指示する
付記12又は13に記載の情報処理装置。
前記予測式集合学習生成手段が
前記操作量決定手段が前記操作量を決定する間隔より長い間隔で前記予測式集合を学習及び生成する
付記1乃至付記14のいずれが1つに記載の情報処理装置。
前記予測式集合学習生成手段が、
前記予測式集合として、
未来の状態変数又は未来の制御変数を従属変数とする複数の関数を含み、
一の関数は、従属変数が未来の制御変数であり、独立変数に未来の状態変数及び未来の制御変数を含まず、
前記一の関数と少なくとも一つの現在以降の操作変数を共通の独立変数とする他の関数を含む
付記1乃至付記15のいずれか1つに記載の情報処理装置。
制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、前記制御対象情報を蓄積し、
前記蓄積された制御対象情報を基に前記制御対象の操作量の決定に用いる予測式集合を学習及び生成し、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記予測式集合と前記蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象を制御に用いる操作量を決定する
予測制御方法。
制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、前記制御対象情報を蓄積する処理と、
前記蓄積された制御対象情報を基に前記制御対象の操作量の決定に用いる予測式集合を学習及び生成する処理と、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記予測式集合と前記蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象を制御に用いる操作量を決定する処理と
をコンピュータに実行させるプログラムを記録したコンピュータ読み取り可能な記録媒体。
制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積する情報蓄積手段と、
前記情報蓄積手段に蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成する予測式集合学習生成手段、および
前記制御対象の操作量の決定に必要な入力情報を受信し、前記入力情報と前記情報蓄積手段に蓄積された制御対象情報と前記予測式集合とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる操作量を決定する操作量決定手段と、
を含む予測制御手段と
あらかじめ入力された式と前記情報蓄積手段に蓄積された制御対象情報とを元に、前記制御対象の制御に用いられる操作量を決定する固定制御手段と
前記操作量決定手段の操作量と前記固定制御手段の操作量のいずれか一つを選択する操作量選択手段と
を含む情報処理装置。
前記予測制御手段及び前記固定制御手段のいずれか、又は、両方を複数含み、
前記操作量選択手段が、
前記一つ又は複数の予測制御手段及び固定制御手段から出力される操作量からいずれか一つを選択する
付記19に記載の情報処理装置。
前記情報蓄積手段が蓄積する制御対象情報、並びに、前記操作量決定手段及び前記固定制御手段の出力する操作量をユーザに表示する表示手段と、
前記操作量選択手段が、ユーザが選択した操作量を出力する
付記19又は付記20に記載の情報処理装置。
前記予測式集合学習生成手段が出力する予測式集合と、前記情報蓄積手段が蓄積する制御対象情報とに基づいて予測値を算出する手段と、
前記予測値と前記情報蓄積手段が蓄積する制御対象情報との差に基づいて予測の誤差を判定する予測誤差判定手段とを含み、
前記操作量選択手段が、前記予測誤差判定手段の結果を基に、前記予測制御手段と前記固定制御手段から出力される操作量からいずれか一つを選択する
付記19又は付記20に記載の情報処理装置。
前記予測制御手段及び前記操作量決定手段が出力する操作量を基に、あらかじめ入力された評価関数に基づいて評価値を計算し、最も評価値の高い制御量を判断する制御評価手段を含み、
前記操作量選択手段が、前記制御評価手段の判断の結果を基に前記予測制御手段と固定制御手段から出力される操作量からいずれか一つを選択する
付記19又は付記20に記載の情報処理装置。
前記制御評価手段が、操作に必要なコストを基に判定する
付記23に記載の情報処理装置。
前記制御評価手段が、前記操作量決定手段の操作量の前記固定制御手段の制御量からの外れの量を基に判定する
付記23に記載の情報処理装置。
前記情報蓄積手段が蓄積する制御対象情報と入力情報の目標値との差に基づいて動作を評価する動作評価手段を含み
前記操作量選択手段が、前記動作評価手段の評価に基づき、前記制御評価手段の結果を基に前記予測制御手段と前記固定制御手段から出力される操作量からいずれか一つを選択する
付記19又は付記20に記載の情報処理装置。
前記予測式集合学習生成部で計算される学習の評価値を基に、評価を行う予測式評価手段を含み、
前記操作量選択手段部が、前記予測式評価手段の結果に基づいて前記予測制御手段及び前記固定制御手段から出力される操作量からいずれか一つを選択する
付記19又は付記20に記載の情報処理装置。
前記予測式評価手段が、学習の尤もらしさを表す情報量基準を基に評価する
付記27に記載の情報処理装置。
前記予測式集合学習生成部より出力される予測式集合が制御可能な予測式か否かを基に、前記予測式集合を評価する予測式集合評価手段を含み
前記操作量選択手段が、前記予測式集合評価手段の結果を基に、前記予測制御手段及び前記固定制御手段から出力される操作量からいずれか一つを選択する
付記19又は付記20に記載の情報処理装置。
前記予測式集合評価手段が、予測式に制御量の変数が含まれる数に基づいて評価する
付記29に記載の情報処理装置。
制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積し、
前記蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成し、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記入力情報と前記蓄積された制御対象情報と前記予測式集合とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる第1の操作量を決定し、
前記第1の操作量を出力し、
あらかじめ入力された式と前記蓄積された制御対象情報とを元に、前記制御対象の制御に用いられる第2の操作量を決定し出力し、
前記第1の操作量と前記第2の操作量のいずれか一つを選択する
予測制御方法。
制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積する処理と、
前記蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成する処理と、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記入力情報と前記蓄積された制御対象情報と前記予測式集合とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる第1の操作量を決定する処理と、
前記第1の操作量を出力する処理と、
あらかじめ入力された式と前記蓄積された制御対象情報とを元に、前記制御対象の制御に用いられる第2の操作量を決定し出力する処理と、
前記第1の操作量と前記第2の操作量のいずれか一つを選択する処理と
をコンピュータに実行させるプログラムを記録したコンピュータ読み取り可能な記録媒体。
101 情報処理システム
102 情報処理システム
600 情報処理装置
610 CPU
620 ROM
630 RAM
640 内部記憶装置
650 IOC
660 入力機器
670 表示機器
680 NIC
700 記憶媒体
2100 制御対象情報
2200 入力情報
2300 操作量
2400 予測値
2500 予測式集合
2501 予測式集合
2600 蓄積情報
2700 計算結果
2800 数理計画問題
2900 予測誤差
3000 判定結果
10000 情報処理装置
10010 情報処理装置
10020 情報処理装置
10030 情報処理装置
11000 情報蓄積部
12000 予測式集合学習生成部
13000 予測式集合格納部
14000 操作量決定部
14100 予測式集合変換部
14200 操作量算出部
14201 操作量算出部
14210 数理計画問題定式化部
14212 制約式生成部
14214 目的関数生成部
14216 許容誤差レジスタ
14217 誤差重み係数レジスタ
14218 コスト係数レジスタ
14220 数理計画問題計算部
14230 操作量良否判定部
14240 操作量選択部
15000 予測値算出部
16000 予測値蓄積部
17000 予測誤差算出部
18000 再学習判定部
20000 周辺環境
21000 制御対象
21001 制御対象
2100n 制御対象
22000 入力装置
23000 出力装置
30000 予測制御部
31000 固定制御部
32000 操作量選択部
33000 表示部
34000 予測誤差判定部
35000 操作量評価部
36000 動作評価部
37000 予測式評価部
Claims (32)
- 制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積する情報蓄積手段と、
前記情報蓄積手段に蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成する予測式集合学習生成手段と、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記予測式集合と前記情報蓄積手段に蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる操作量を決定する操作量決定手段と、
含む情報処理装置。 - 前記予測式集合が、
未来の予測制御モデルの状態を表す状態変数又は未来の予測制御モデルの制御対象となる制御変数を従属変数とする二つ以上の予測式を含み、
前記予測式として、
独立変数が現在以降の予測制御モデルの操作を表す操作変数を含み、未来の状態変数及び未来の制御変数を含まず、従属変数が未来の制御変数である第一の予測式と、
少なくとも一つの現在以降の予測制御モデルの操作を表す操作変数を前記第一の予測式と共通の独立変数とする第二の予測式と
を含む請求項1に記載の情報処理装置。 - 前記操作量決定手段が、
前記入力情報として、前記予測制御モデルにおいて制御の対象となる制御変数に対する設定値又は制約のいずれか又は両方を受信する
請求項1又は2に記載の情報処理装置。 - 前記操作量決定手段が
前記入力情報として、
さらに、複数の前記制御変数における重要性を表す優先度を受信する
請求項3に記載の情報処理装置。 - 前記予測集合学習生成手段が生成した予測式集合を格納する予測式集合格納手段をさらに含み、
前記操作量決定手段が、
前記制御対象への操作量を決定に、前記予測式集合格納手段に格納された予測式集合を用いる
請求項1乃至請求項4のいずれか1項に記載の情報処理装置。 - 前記予測式集合格納手段に格納された予測式集合と、前記受信した制御対象情報と、前記操作量とを基に、前記制御対象の予測値を算出する予測値算出手段をさらに含む
請求項1乃至請求項5のいずれか1項に記載の情報処理装置。 - 前記予測値算出手段が、
前記予測値の算出に前記情報蓄積手段に蓄積された制御対象情報を用いる
請求項6に記載の情報処理装置。 - 前記予測値を蓄積する予測値蓄積手段と、
前記予測値蓄積手段に蓄積された予測値と前記情報蓄積手段に蓄積された制御対象情報とを基に予測誤差を算出する予測誤差算出手段と、
前記予測誤差を用いて前記予測式集合学習生成手段において再学習するか否か判定する再学習判定手段と
をさらに含み、
前記予測式集合学習生成手段が、
前記再学習判定手段の判定結果に基づいて前記予測式集合を再学習及び再生成する
請求項6又は請求項7に記載の情報処理装置。 - 前記情報蓄積手段が、複数の制御対象情報を受信し、蓄積し、
前記操作量決定手段が、前記複数の制御対象の操作量を決定する
請求項1乃至8のいずれか1項に記載の情報処理装置。 - 前記操作量決定手段が、
前記予測式集合を所定の規則に従い変換する予測式集合変換手段と、
前記予測式集合変換手段で変換された予測式集合と前記情報蓄積手段に蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象への操作量を算出する操作量算出手段と
を含む請求項1乃至9のいずれか1項に記載の情報処理装置。 - 前記操作量算出手段が、
前記予測式集合変換手段で変換された予測式集合と前記情報蓄積手段に蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に前記予測制御モデルを表す数理計画問題を定式化する数理計画問題定式化手段と、
定式化された前記数理計画問題を計算し、計算結果を前記操作量と決定する数理計画問題計算手段と
を含む請求項10に記載の情報処理装置。 - 前記数理計画問題計算手段の計算結果の良否を判定する操作量良否判定手段と、
前記操作量良否判定手段の判定結果が良の場合に前記計算結果を前記操作量と決定する操作量選択手段とを含む請求項11に記載の情報処理装置。 - 前記操作量選択手段が、
前記操作量判定手段が複数の計算結果を良と判定した場合に、
前記入力情報又は所定の規則を基に前記計算結果から前記操作量を選択する
請求項12に記載の情報処理装置。 - 前記操作量良否判定手段が
前記計算結果をすべて否と判定した場合に、前記数理計画問題定式化手段に再度の定式化を指示する
請求項12又は13に記載の情報処理装置。 - 前記予測式集合学習生成手段が
前記操作量決定手段が前記操作量を決定する間隔より長い間隔で前記予測式集合を学習及び生成する
請求項1乃至請求項14のいずれが1項に記載の情報処理装置。 - 前記予測式集合学習生成手段が、
前記予測式集合として、
未来の状態変数又は未来の制御変数を従属変数とする複数の関数を含み、
一の関数は、従属変数が未来の制御変数であり、独立変数に未来の状態変数及び未来の制御変数を含まず、
前記一の関数と少なくとも一つの現在以降の操作変数を共通の独立変数とする他の関数を含む
請求項1乃至請求項15のいずれか1項に記載の情報処理装置。 - 制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、前記制御対象情報を蓄積し、
前記蓄積された制御対象情報を基に前記制御対象の操作量の決定に用いる予測式集合を学習及び生成し、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記予測式集合と前記蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象を制御に用いる操作量を決定する
予測制御方法。 - 制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、前記制御対象情報を蓄積する処理と、
前記蓄積された制御対象情報を基に前記制御対象の操作量の決定に用いる予測式集合を学習及び生成する処理と、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記予測式集合と前記蓄積された制御対象情報と前記受信した制御対象情報と前記入力情報とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象を制御に用いる操作量を決定する処理と
をコンピュータに実行させるプログラムを記録したコンピュータ読み取り可能な記録媒体。 - 制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積する情報蓄積手段と、
前記情報蓄積手段に蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成する予測式集合学習生成手段、および、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記入力情報と前記情報蓄積手段に蓄積された制御対象情報と前記予測式集合とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる第1の操作量を決定する操作量決定手段と、
を含み前記第1の操作量を出力する予測制御手段と
あらかじめ入力された式と前記情報蓄積手段に蓄積された制御対象情報とを元に、前記制御対象の制御に用いられる第2の操作量を決定し出力する固定制御手段と、
前記操作量決定手段の第1の操作量と前記固定制御手段の第2の操作量のいずれか一つを選択する操作量選択手段と
を含む情報処理装置。 - 前記予測制御手段及び前記固定制御手段のいずれか、又は、両方を複数含み、
前記操作量選択手段が、
前記一つ又は複数の予測制御手段から出力される第1の操作量及び固定制御手段から出力される第2の操作量の中からいずれか一つを選択する
請求項19に記載の情報処理装置。 - 前記情報蓄積手段が蓄積する制御対象情報、並びに、前記操作量決定手段及び前記固定制御手段の出力する操作量をユーザに表示する表示手段と、
前記操作量選択手段が、前記予測制御手段から出力される第1の操作量及び前記固定制御手段から出力される第2の操作量の中からユーザが選択した操作量を選択して出力する
請求項19又は請求項20に記載の情報処理装置。 - 前記予測式集合学習生成手段が出力する予測式集合と、前記情報蓄積手段が蓄積する制御対象情報とに基づいて予測値を算出する手段と、
前記予測値と前記情報蓄積手段が蓄積する制御対象情報との差に基づいて予測の誤差を判定する予測誤差判定手段とを含み、
前記操作量選択手段が、前記予測誤差判定手段の結果を基に、前記予測制御手段から出力される第1の操作量及び前記固定制御手段から出力される第2の操作量の中からいずれか一つを選択する
請求項19又は請求項20に記載の情報処理装置。 - 前記予測制御手段及び前記操作量決定手段が出力する操作量を基に、あらかじめ入力された評価関数に基づいて評価値を計算し、最も評価値の高い制御量を判断する制御評価手段を含み、
前記操作量選択手段が、前記制御評価手段の判断の結果を基に前記予測制御手段から出力される第1の操作量及び前記固定制御手段から出力される第2の操作量の中からいずれか一つを選択する
請求項19又は請求項20に記載の情報処理装置。 - 前記制御評価手段が、操作に必要なコストを基に判定する
請求項23に記載の情報処理装置。 - 前記制御評価手段が、前記操作量決定手段から出力される第1の操作量の前記固定制御手段の制御量からの外れの量を基に判定する
請求項23に記載の情報処理装置。 - 前記情報蓄積手段が蓄積する制御対象情報と入力情報の目標値との差に基づいて動作を評価する動作評価手段を含み
前記操作量選択手段が、前記動作評価手段の評価に基づき、前記制御評価手段の結果を基に前記予測制御手段から出力される第1の操作量と前記固定制御手段から出力される第2の操作量の中からいずれか一つを選択する
請求項19又は請求項20に記載の情報処理装置。 - 前記予測式集合学習生成手段で計算される学習の評価値を基に、評価を行う予測式評価手段を含み、
前記操作量選択手段が、前記予測式評価手段の結果に基づいて前記予測制御手段から出力される第1の操作量及び前記固定制御手段から出力される第2の操作量の中からいずれか一つを選択する
請求項19又は請求項20に記載の情報処理装置。 - 前記予測式評価手段が、学習の尤もらしさを表す情報量基準を基に評価する
請求項27に記載の情報処理装置。 - 前記予測式集合学習生成手段より出力される予測式集合が制御可能な予測式か否かを基に、前記予測式集合を評価する予測式集合評価手段を含み
前記操作量選択手段が、前記予測式集合評価手段の結果を基に、前記予測制御手段から出力される第1の操作量及び前記固定制御手段から出力される第2の操作量の中からいずれか一つを選択する
請求項19又は請求項20に記載の情報処理装置。 - 前記予測式集合評価手段が、予測式に制御量の変数が含まれる数に基づいて評価する
請求項29に記載の情報処理装置。 - 制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積し、
前記蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成し、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記入力情報と前記蓄積された制御対象情報と前記予測式集合とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる第1の操作量を決定し、
前記第1の操作量を出力し、
あらかじめ入力された式と前記蓄積された制御対象情報とを元に、前記制御対象の制御に用いられる第2の操作量を決定し出力し、
前記第1の操作量と前記第2の操作量のいずれか一つを選択する
予測制御方法。 - 制御対象及び前記制御対象を含む周辺環境に関する情報を含む制御対象情報を受信し、蓄積する処理と、
前記蓄積された制御対象情報を基に前記制御対象に対する操作量の決定に用いる予測式集合を学習及び生成する処理と、
前記制御対象の操作量の決定に必要な入力情報を受信し、前記入力情報と前記蓄積された制御対象情報と前記予測式集合とを基に、前記制御対象の予測制御モデルを構築して、前記制御対象の制御に用いられる第1の操作量を決定する処理と、
前記第1の操作量を出力する処理と、
あらかじめ入力された式と前記蓄積された制御対象情報とを元に、前記制御対象の制御に用いられる第2の操作量を決定し出力する処理と、
前記第1の操作量と前記第2の操作量のいずれか一つを選択する処理と
をコンピュータに実行させるプログラムを記録したコンピュータ読み取り可能な記録媒体。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/914,761 US10048658B2 (en) | 2013-09-12 | 2014-05-08 | Information processing device, predictive control method, and recording medium |
JP2015536422A JPWO2015037165A1 (ja) | 2013-09-12 | 2014-05-08 | 情報処理装置、予測制御方法、及び、プログラム |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2013-189500 | 2013-09-12 | ||
JP2013189500 | 2013-09-12 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2015037165A1 true WO2015037165A1 (ja) | 2015-03-19 |
Family
ID=52665302
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2014/002441 WO2015037165A1 (ja) | 2013-09-12 | 2014-05-08 | 情報処理装置、予測制御方法及び記録媒体 |
Country Status (3)
Country | Link |
---|---|
US (1) | US10048658B2 (ja) |
JP (1) | JPWO2015037165A1 (ja) |
WO (1) | WO2015037165A1 (ja) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2017030137A (ja) * | 2015-07-31 | 2017-02-09 | ファナック株式会社 | 人の行動パターンを学習する機械学習装置、ロボット制御装置、ロボットシステム、および機械学習方法 |
JP2020194453A (ja) * | 2019-05-29 | 2020-12-03 | メタウォーター株式会社 | モデル予測制御システム、情報処理装置、プログラム、及びモデル予測制御方法 |
JP2020197967A (ja) * | 2019-06-04 | 2020-12-10 | メタウォーター株式会社 | モデル予測制御システム、情報処理装置、プログラム、及びモデル予測制御方法 |
US11216733B2 (en) | 2017-11-20 | 2022-01-04 | Electronics And Telecommunications Research Institute | Self-evolving agent-based simulation system and method thereof |
JP2022065773A (ja) * | 2020-10-16 | 2022-04-28 | 横河電機株式会社 | 制御装置、コントローラ、制御システム、制御方法、および制御プログラム |
US11400954B2 (en) | 2016-06-09 | 2022-08-02 | Nec Corporation | Vehicle control system, vehicle control method, and program recording medium |
US11780095B2 (en) | 2015-07-31 | 2023-10-10 | Fanuc Corporation | Machine learning device, robot system, and machine learning method for learning object picking operation |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6517762B2 (ja) * | 2016-08-23 | 2019-05-22 | ファナック株式会社 | 人とロボットが協働して作業を行うロボットの動作を学習するロボットシステム |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH05120256A (ja) * | 1991-10-29 | 1993-05-18 | Toshiba Corp | モデル予測制御装置 |
JPH05303406A (ja) * | 1992-04-28 | 1993-11-16 | Hitachi Ltd | 定性推論予測制御方法 |
JPH05324007A (ja) * | 1992-05-18 | 1993-12-07 | Yokogawa Electric Corp | プロセス制御装置 |
JPH1153005A (ja) * | 1997-08-04 | 1999-02-26 | Nippon Steel Corp | 制御系におけるモデルパラメータの学習方法 |
JP2005118959A (ja) * | 2003-10-17 | 2005-05-12 | Toyoda Mach Works Ltd | 作業支援装置、作業支援方法、位置決め作業支援装置およびパワーアシスト作業支援装置 |
JP2011118786A (ja) * | 2009-12-04 | 2011-06-16 | Sony Corp | 情報処理装置、観測値予測方法、及びプログラム |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4663703A (en) * | 1985-10-02 | 1987-05-05 | Westinghouse Electric Corp. | Predictive model reference adaptive controller |
US5519605A (en) * | 1994-10-24 | 1996-05-21 | Olin Corporation | Model predictive control apparatus and method |
EP2297618B1 (en) * | 2008-06-12 | 2019-12-11 | Valmet Automation, Inc. | Method and apparatus for reel building and roll runnability in moving web manufacturing |
US8504175B2 (en) * | 2010-06-02 | 2013-08-06 | Honeywell International Inc. | Using model predictive control to optimize variable trajectories and system control |
US8711211B2 (en) * | 2010-06-14 | 2014-04-29 | Howard Hughes Medical Institute | Bessel beam plane illumination microscope |
-
2014
- 2014-05-08 JP JP2015536422A patent/JPWO2015037165A1/ja active Pending
- 2014-05-08 WO PCT/JP2014/002441 patent/WO2015037165A1/ja active Application Filing
- 2014-05-08 US US14/914,761 patent/US10048658B2/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH05120256A (ja) * | 1991-10-29 | 1993-05-18 | Toshiba Corp | モデル予測制御装置 |
JPH05303406A (ja) * | 1992-04-28 | 1993-11-16 | Hitachi Ltd | 定性推論予測制御方法 |
JPH05324007A (ja) * | 1992-05-18 | 1993-12-07 | Yokogawa Electric Corp | プロセス制御装置 |
JPH1153005A (ja) * | 1997-08-04 | 1999-02-26 | Nippon Steel Corp | 制御系におけるモデルパラメータの学習方法 |
JP2005118959A (ja) * | 2003-10-17 | 2005-05-12 | Toyoda Mach Works Ltd | 作業支援装置、作業支援方法、位置決め作業支援装置およびパワーアシスト作業支援装置 |
JP2011118786A (ja) * | 2009-12-04 | 2011-06-16 | Sony Corp | 情報処理装置、観測値予測方法、及びプログラム |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2017030137A (ja) * | 2015-07-31 | 2017-02-09 | ファナック株式会社 | 人の行動パターンを学習する機械学習装置、ロボット制御装置、ロボットシステム、および機械学習方法 |
US10807235B2 (en) | 2015-07-31 | 2020-10-20 | Fanuc Corporation | Machine learning device, robot controller, robot system, and machine learning method for learning action pattern of human |
US11780095B2 (en) | 2015-07-31 | 2023-10-10 | Fanuc Corporation | Machine learning device, robot system, and machine learning method for learning object picking operation |
US11904469B2 (en) | 2015-07-31 | 2024-02-20 | Fanuc Corporation | Machine learning device, robot controller, robot system, and machine learning method for learning action pattern of human |
US11400954B2 (en) | 2016-06-09 | 2022-08-02 | Nec Corporation | Vehicle control system, vehicle control method, and program recording medium |
US11216733B2 (en) | 2017-11-20 | 2022-01-04 | Electronics And Telecommunications Research Institute | Self-evolving agent-based simulation system and method thereof |
JP2020194453A (ja) * | 2019-05-29 | 2020-12-03 | メタウォーター株式会社 | モデル予測制御システム、情報処理装置、プログラム、及びモデル予測制御方法 |
JP7287835B2 (ja) | 2019-05-29 | 2023-06-06 | メタウォーター株式会社 | モデル予測制御システム、情報処理装置、プログラム、及びモデル予測制御方法 |
JP2020197967A (ja) * | 2019-06-04 | 2020-12-10 | メタウォーター株式会社 | モデル予測制御システム、情報処理装置、プログラム、及びモデル予測制御方法 |
JP7353804B2 (ja) | 2019-06-04 | 2023-10-02 | メタウォーター株式会社 | モデル予測制御システム、情報処理装置、プログラム、及びモデル予測制御方法 |
JP2022065773A (ja) * | 2020-10-16 | 2022-04-28 | 横河電機株式会社 | 制御装置、コントローラ、制御システム、制御方法、および制御プログラム |
JP7342833B2 (ja) | 2020-10-16 | 2023-09-12 | 横河電機株式会社 | 制御装置、コントローラ、制御システム、制御方法、および制御プログラム |
Also Published As
Publication number | Publication date |
---|---|
US10048658B2 (en) | 2018-08-14 |
JPWO2015037165A1 (ja) | 2017-03-02 |
US20160209817A1 (en) | 2016-07-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2015037165A1 (ja) | 情報処理装置、予測制御方法及び記録媒体 | |
Mustafaraj et al. | Prediction of room temperature and relative humidity by autoregressive linear and nonlinear neural network models for an open office | |
Li et al. | Predication control for indoor temperature time-delay using Elman neural network in variable air volume system | |
Kumar et al. | A novel hybrid model based on particle swarm optimisation and extreme learning machine for short-term temperature prediction using ambient sensors | |
EP3200038A1 (en) | Model evaluation device, model evaluation method, and program recording medium | |
Zou et al. | Model predictive control based on particle swarm optimization of greenhouse climate for saving energy consumption | |
Yiu et al. | Multiple ARMAX modeling scheme for forecasting air conditioning system performance | |
Özbek et al. | Design of an optimal fractional fuzzy gain-scheduled Smith Predictor for a time-delay process with experimental application | |
CN104049598B (zh) | 线性和非线性系统的基于确定性优化的控制系统及方法 | |
Han et al. | A review of reinforcement learning methodologies on control systems for building energy | |
Ellis | Machine learning enhanced grey-box modeling for building thermal modeling | |
Di Natale et al. | Lessons learned from data-driven building control experiments: Contrasting gaussian process-based mpc, bilevel deepc, and deep reinforcement learning | |
Homod et al. | Deep clustering of Lagrangian trajectory for multi-task learning to energy saving in intelligent buildings using cooperative multi-agent | |
Stoffel et al. | Real-life data-driven model predictive control for building energy systems comparing different machine learning models | |
Wang et al. | Globally optimal nonlinear model predictive control based on multi-parametric disaggregation | |
WO2016203757A1 (ja) | 制御装置、それを使用する情報処理装置、制御方法、並びにコンピュータ・プログラムが格納されているコンピュータ読み取り可能な記憶媒体 | |
Revati et al. | Facilitating energy-efficient operation of smart building using data-driven approaches | |
Lee et al. | On-policy learning-based deep reinforcement learning assessment for building control efficiency and stability | |
Kontes et al. | Adaptive-fine tuning of building energy management systems using co-simulation | |
Tan et al. | Performance-based control system design automation via evolutionary computing | |
Schepers et al. | Autonomous building control using offline reinforcement learning | |
Ji et al. | Identification and predictive control for a circulation fluidized bed boiler | |
Dey et al. | Reinforcement Learning Building Control: An Online Approach With Guided Exploration Using Surrogate Models | |
JP7060130B1 (ja) | 運用支援装置、運用支援方法及びプログラム | |
Javed et al. | Modelling and optimization of residential heating system using random neural networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 14844330 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2015536422 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14914761 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 14844330 Country of ref document: EP Kind code of ref document: A1 |