WO2002069257A1 - Dispositif de recherche de solution optimale, dispositif pour commander un objet commande par algorithme d'optimisation, et programme de recherche de solution optimale - Google Patents
Dispositif de recherche de solution optimale, dispositif pour commander un objet commande par algorithme d'optimisation, et programme de recherche de solution optimale Download PDFInfo
- Publication number
- WO2002069257A1 WO2002069257A1 PCT/JP2002/000932 JP0200932W WO02069257A1 WO 2002069257 A1 WO2002069257 A1 WO 2002069257A1 JP 0200932 W JP0200932 W JP 0200932W WO 02069257 A1 WO02069257 A1 WO 02069257A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- evaluation
- control
- evaluation function
- output
- individual
- Prior art date
Links
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
- F02D41/1405—Neural network control
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
-
- 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
-
- 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
- G05B13/027—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 using neural networks only
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
- F02D41/1406—Introducing closed-loop corrections characterised by the control or regulation method with use of a optimisation method, e.g. iteration
-
- 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
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32028—Electronic catalog, to select material, resources, make lists with prices
-
- 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
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32393—Host and central distribution control between storage and cells
Definitions
- the present invention relates to an apparatus and a program for searching for an optimum solution of an evaluation function by an optimization algorithm, and in particular, to search for an optimum solution according to a user's desire, and to reduce the time required for the optimization.
- the present invention relates to an optimal solution search device, a control device of a control target by an optimization algorithm, and an optimal solution search program.
- the characteristics of the products to be controlled are assumed to be those who will use the products at the development and design stages, and the virtual users Taking into account the tastes and usage conditions of the users, it is decided to adapt to the widest possible range of users.
- the users who use the above products have their own unique personalities, and their preferences vary widely, so as described above, the preferences of the users who are likely to use the products, etc. Even if the product is developed and designed under the assumption that it is, it is almost impossible to provide characteristics that are satisfied by all users.
- a control method that estimates the user's preferences and usage conditions after purchase using an optimization algorithm such as GA, and changes the control characteristics to characteristics that the user can satisfy, has been developed. Attempted.
- control results are quantified and the control characteristics are evaluated autonomously based on predetermined evaluation criteria.
- Autonomous evaluation method that optimizes control characteristics
- interactive evaluation method that optimizes control characteristics while displaying the control results to the user and evaluating through dialogue with the user.
- the autonomous evaluation method is performed autonomously based on a predetermined evaluation criterion, the control characteristics can be optimized in a relatively short time.However, an evaluation criterion that optimizes the user's preference is used. It is not suitable when it changes from person to person. On the other hand, the interactive evaluation method is suitable when the evaluation criterion that optimizes the user's preference changes depending on the user, but the evaluation by the user is much more rapid. Since it is not possible to obtain, for example, if all the control characteristics of the vehicle are to be optimized by this method, there is a problem that it takes a relatively long time to complete the optimization.
- the present invention has been made by focusing on such unresolved problems of the conventional technology, and it is possible to search for an optimal solution according to a user's desire, and furthermore, it is necessary for optimization. It is an object of the present invention to provide an optimal solution search device capable of reducing time, a control device of a control target by an optimization algorithm, and an optimal solution search program. Disclosure of the invention
- evaluation is performed by a combination of interactive evaluation and autonomous evaluation. While the output of the valence function is repeatedly evaluated, the optimal solution of the valuation function is searched by the optimization algorithm.
- the invention described in claim 1 can be applied, for example, when optimizing the control characteristics of a control target.
- the invention is not limited to this, and various optimization algorithms to which the optimization algorithm has been applied have been applied.
- the problem can also be applied.
- the invention described in claim 1 can employ an evolutionary optimization algorithm such as GA, GP (Genetic Progress) or ES (evolutional Stratage) as the optimization algorithm. It is also possible to use a type optimization algorithm or an optimization algorithm based on reinforcement learning. Hereinafter, the same applies to a control device to be controlled by the optimization algorithm according to claim 3 and an optimal solution search program according to claim 1'2.
- GA Genetic Progress
- ES evolutional Stratage
- the output of the evaluation function is calculated as the evaluation value by the evaluation value calculation means, and the calculated evaluation value is stored in the storage means.
- the interactive evaluation is performed based on the input contents.
- autonomous evaluation is performed based on the evaluation value of the storage means. And these interactive and autonomous evaluations While the output of the evaluation function is repeatedly evaluated by the combination of, the optimal solution of the evaluation function is searched for by the optimization algorithm.
- the storage means stores the evaluation value by any means and at all times, and may store the evaluation value in advance, or may store the evaluation value without storing the evaluation value in advance.
- the evaluation value may be stored by an external input or the like.
- a control device of a control target by an optimization algorithm according to claim 3 of the present invention has an effect on control characteristics of a control system that controls characteristics of the control target.
- An evaluation function that outputs the control characteristic with a control coefficient as an input searching for the optimal solution of the evaluation function by an optimization algorithm while repeatedly evaluating the output of the evaluation function, the control characteristic of the control system is obtained.
- An apparatus for optimizing comprising: an interactive evaluation for evaluating an output of the evaluation function based on a dialog with a user; and an autonomous evaluation for evaluating an output of the evaluation function based on a predetermined evaluation criterion. The search by the optimization algorithm is performed in combination with the above.
- the optimal solution of the evaluation function is searched for by the optimization algorithm.
- the control system operates based on the control coefficient searched as the optimal solution, thereby optimizing the control characteristics of the control system.
- control device of the controlled object by the optimization algorithm according to claim 4 of the present invention is the control device of the controlled object by the optimization algorithm according to claim 3, wherein the evaluation function For storing the output of A storage unit, an evaluation value calculation unit that calculates an output of the evaluation function as an evaluation value and stores the evaluation value in the storage unit, and an evaluation input unit that inputs an evaluation by the user, wherein the interactive evaluation includes: The autonomous evaluation is performed based on the evaluation content of the storage means, based on the input content of the evaluation input means.
- a control device for a controlled object by the optimization algorithm according to claim 5 of the present invention is a control device for a controlled object by the optimization algorithm according to any one of claims 3 and 4.
- the apparatus comprises: a first control system that controls a first characteristic of the control target; and a second control system that controls a second characteristic of the control target, wherein the first control system includes the first control system.
- the output of the first evaluation function is repeatedly evaluated by the interactive evaluation using the first evaluation function that outputs the control characteristic with a control coefficient that affects the control characteristic of the control system as the input.
- An optimal solution of the function is searched for by the optimization algorithm.
- For the second control system a control coefficient that affects a control characteristic of the second control system is input and the control characteristic is output.
- the first evaluation function is within a predetermined range based on an output of the first evaluation function obtained from a solution searched as an optimal solution in the optimization of the first control system. While repeatedly evaluating the output of the second evaluation function by the autonomous evaluation so that the output of An optimal solution of the second evaluation function is searched for by the optimization algorithm.
- the optimal solution of the first evaluation function is searched for by the optimization algorithm while the output of the first evaluation function is repeatedly evaluated by interactive evaluation. Then, for the second control system, the output of the first evaluation function belongs to a predetermined range based on the output of the first evaluation function obtained from the solution searched as the optimal solution in the optimization of the first control system, and While the output of the second evaluation function is repeatedly evaluated by the autonomous evaluation, the optimal solution of the second evaluation function is searched by the optimization algorithm.
- the first evaluation function is within a predetermined range based on an output of the first evaluation function obtained from a solution searched as an optimal solution in the optimization of the first control system.
- the optimal solution of the second evaluation function is searched by the optimization algorithm while repeatedly evaluating the output of the second evaluation function by the interactive evaluation so that the output of the second evaluation function belongs.
- a control device for a controlled object by an optimization algorithm according to claim 7 of the present invention is a control device for a controlled object by an optimization algorithm according to any one of claims 5 and 6.
- the controlled object is an engine
- those to be subjected to the interactive evaluation are a rotation speed change rate and a throttle of the engine.
- the response which is determined by the rate of change of the opening degree and which is subject to the autonomous evaluation among the outputs of the first evaluation function and the second evaluation function is the fuel efficiency of the engine.
- the response characteristics of the first evaluation function and the second evaluation function to be subjected to interactive evaluation are optimized through repeated interactive evaluations.
- those subject to autonomous evaluation are repeatedly subjected to autonomous evaluation to optimize the fuel efficiency characteristics of the engine.
- the individual information operation means performs an information operation imitating the genetic operation on the individual information
- the evaluation value calculation means calculates the evaluation value
- the individual selection means calculates the evaluation value.
- the survival or selection of individuals is performed based on the evaluation value.
- the generation proceeds by performing the genetic operation by the individual information operation means and the individual selection operation by the individual selection means at least once in the same generation.
- control device of the controlled object by the optimization algorithm according to claim 9 is the control device of the controlled object by the optimization algorithm according to claim 8, Is an engine, wherein the individual information includes, as the control coefficient, a fuel injection amount of the engine, a transient correction amount for correcting the fuel injection amount in a transient state of the engine, a correction value of the fuel injection amount, or The correction value of the transient correction amount is assigned.
- the fuel injection amount, the transient correction amount, the correction value of the fuel injection amount or the transient value A correction value of the correction amount is determined.
- a control device of a control target by an optimization algorithm according to claim 10 of the present invention is a control device of a control target by an optimization algorithm according to claim 8, wherein the control target is A fuel injection amount of the engine, a transient correction amount for correcting the fuel injection amount in a transient state of the engine, a correction value of the fuel injection amount, or the transient correction.
- a correction value of the amount is generated by a neural network, and the individual information is assigned a synaptic coupling coefficient in the neural network as the control coefficient.
- the fuel injection amount, the transient correction amount, the correction value of the fuel injection amount, or the correction value of the transient correction amount are generated by the neural network, but the population is increased in the direction of improving the evaluation value. As it evolves, the synaptic coupling coefficient in the neural network that can be expected to obtain high evaluation values is determined.
- a control device for a control target by the optimization algorithm according to claim 11 of the present invention is a control device for a control target by the optimization algorithm according to any one of claims 5 and 6.
- the control target is an electric motor
- a target of the interactive evaluation among the first evaluation function and the second evaluation function output is a rotation change of the electric motor.
- the output of the first evaluation function and the output of the second evaluation function to be subjected to the autonomous evaluation is the power consumption of the electric motor.
- an optimal solution search program provides an optimal solution of the above-mentioned evaluation function while repeatedly evaluating the output of the evaluation function.
- a computer-executable program that performs an evaluation on an output of the evaluation function based on a dialog with a user, and an evaluation on an output of the evaluation function based on a predetermined evaluation criterion.
- This is a program that causes a computer to execute a search process using the optimization algorithm.
- FIG. 1 is a diagram showing the basic concept of the present invention.
- FIG. 2 is a block diagram showing a basic configuration of the present invention.
- FIG. 3 is a block diagram showing a configuration of an engine control system to which the present invention is applied.
- FIG. 4 is a block diagram showing the configuration of the interactive optimization unit 310.
- FIG. 5 is a diagram showing a data structure of individual information.
- FIG. 6 is a flowchart showing the processing executed by the evolution adaptation unit 330.
- FIG. 7 is a block diagram showing the configuration of the autonomous optimization unit 350.
- FIG. 8 is a diagram showing the configuration of the neural networks 36 2 a and 36 2 b and the data structure of the individual information.
- FIG. 9 is a flowchart showing the processing executed by the evolution adaptation unit 360.
- FIG. 10 is a diagram showing an order for optimizing the control characteristics of the engine 10 and the electronic throttle 12.
- FIG. 11 is a diagram conceptually showing a state in which the teacher data set acquires a new teacher data.
- FIG. 12 is a diagram conceptually showing the update of the teacher data set.
- FIGS. 1 to 12 are diagrams showing an embodiment of an optimum solution search device, a control device to be controlled by an optimization algorithm, and an optimum solution search program according to the present invention.
- FIG. Figure 1 shows the It is a figure showing the basic concept of light.
- the basic configuration of the present invention is based on an optimization target 1 and a first operation amount of the optimization target 1 by an evolutionary optimization algorithm while evaluating the optimization target 1 by interactive evaluation.
- An interactive optimization unit 2 that determines and outputs the optimization target, and an autonomous optimization that determines and outputs the second operation amount of the optimization target 1 using an evolutionary optimization algorithm while evaluating the optimization target 1 by an autonomous evaluation 3
- the interactive optimization unit 2 determines, based on the operation result of the optimization target 1, a first operation amount that optimizes the operation characteristics of the optimization target 1 by using the GA, and determines the determined first operation amount. Is output to optimization target 1.
- the autonomous optimization unit 3 determines, based on the operation result of the optimization target 1, a second operation amount that optimizes the operation characteristics of the optimization target 1 by using the GA, and determines the determined second operation amount. Is output to optimization target 1.
- FIG. 2 is a block diagram showing a basic configuration of the present invention.
- the basic configuration of the present invention includes a control target 50, and a control device 60 that controls a control amount of the control target 50 based on a control result of the control target 50. .
- the control device 60 is composed of three control layers: a reflection layer 500, a learning layer 600, and an evolution adaptive layer 700, and receives a control result from the control target 50, and based on the input control result.
- the reflection layer 500 determines the basic control amount
- the learning layer 600 and the evolution adaptive layer 700 determine the correction rate for the basic control amount
- the final control amount is determined from the basic control amount and the correction rate.
- the configurations of the reflection layer 500, the learning layer 600, and the evolution adaptive layer 700 will be described in detail.
- the reflection layer 500 is composed of a basic control amount and a control connection in the form of a mathematical expression, a map, a neural network, a file, a subsample architecture, and the like.
- the basic control unit 510 defines the relationship with the control result.
- the basic control unit 510 inputs the control result from the control target 50, determines the basic control amount based on the input control result, and outputs it. I do.
- the subsampling architecture is known as behavioral artificial intelligence that performs parallel processing.
- the evolutionary adaptation layer 7100 has an interactive optimization unit 710 that optimizes the control characteristics of the control target 50 using an evolutionary optimization algorithm while repeatedly performing interactive evaluations. It is composed of an autonomous optimization unit 750 that optimizes the control characteristics of the control target 50 using an evolutionary optimization algorithm.
- the interactive optimization unit 7100 has at least one control module that outputs the control amount of the control target 50 based on the control result, and evaluates the control characteristics of the control target 50 with the user in an interactive manner.
- the c GA is configured to optimize the control module by GA, to generate a population consisting of a set of a plurality of individuals virtually, each individual Each time, the individual information is configured based on the genetic information of the individual.
- a control coefficient for constructing a control module is assigned to each individual information c.
- a genetic operation that simulates a genetic operation is performed on the individual information, and an evaluation value of the individual.
- the autonomous optimization unit 750 uses the base from the reflection layer 500 based on the control result. It has at least one control module that outputs an evolutionary correction factor for correcting this control amount to a value that meets the user's wishes, and is configured to optimize the control module by GA. After the optimal control module is constructed, the control module of the autonomous optimization unit 750 is fixed to the optimal control module, and control is performed using the evolution correction rate that corrects the basic control amount from the reflective layer 500.
- the learning layer 600 learns information about the optimal control module. After the information on the optimal control module is learned by the learning layer 600, the output is returned to “1”, and thereafter, the operation is performed according to the instruction of the user. That is, the control by the control module of the autonomous optimization unit 750 is performed only during the evolution simulation and during the learning.
- the learning layer 600 includes a learning unit 6100 having two neural networks that can be switched between learning and execution, and the learning unit 6100 is controlled by one dual network (for execution). While performing, the other neural network (for learning) learns the relationship between the input and output of the optimal control module from the evolutionary adaptation layer 700.
- the neural network performing control and the neural network after learning are switched, and the control module obtained from the learning result in the neural network after learning is used.
- the control starts, and the neural network that was executing the control begins to function for learning.
- the neural network in the learning layer 600 is set so as to output “1” in the initial state. Therefore, in the initial state, the reflection layer 500 and the evolution adaptive layer are set in the initial state. Control according to 700 is performed.
- the execution neural network is configured to receive a control result from the control target 500 and output a learning correction rate for correcting the basic control amount from the reflection layer 500 based on the input control result. .
- This configuration is the same for the learning neural network.
- the control device 600 adds the learning correction rate from the learning layer 600 and the evolution correction rate from the evolution adaptive layer 700, and adds the result to the basic control amount from the reflection layer 500. To calculate the control amount. This control amount is output to the control target 50.
- the optimal solution search device, the control device to be controlled by the optimization algorithm, and the optimal solution search program according to the present invention are based on a combination of interactive evaluation and autonomous evaluation. This was applied to the case where the fuel economy characteristics and response characteristics of the engine 10 were optimized for the user by GA while repeating the evaluation.
- FIG. 3 is a block diagram showing a configuration of an engine control system to which the present invention is applied.
- the engine control system detects the operating state of the engine 10, the electronic throttle 12 that adjusts the intake air amount to the cylinder, and the engine 1 and the electronic throttle 12 to detect the operating state of the engine 1.
- the various sensors 20 that output various kinds of information (hereinafter, collectively referred to as outside world information) relating to the operating state of 0, and a control device that controls the fuel injection amount of the engine 10 based on the outside world information from the various kinds of sensors 20 30.
- the various sensors 20 detect the operating state of the engine 10 and the electronic throttle 12 and the running state of the vehicle, and based on the detection results, the rotation speed of the engine 10, the throttle opening, and the rate of change of the throttle opening.
- the distance pulse and the fuel injection amount are output as external information.
- the control device 30 is composed of three control layers: a reflection layer 100, a learning layer 200, and an evolution adaptation layer 300, and inputs external information from various sensors 20 and based on the input external information.
- the reflective layer 100 determines the basic fuel injection amount, and the learning layer
- the correction amount for the basic injection amount is determined in 200 and the evolution adaptive layer 300, and the final fuel injection amount is determined from the basic injection amount and the correction amount.
- the configurations of the reflective layer 100, the learning layer 200, and the evolution adaptive layer 300 will be described in detail.
- the reflection layer 100 is a basic control unit that defines the relationship between the basic injection amount, the transient correction rate, and the external information in the form of a mathematical expression, a map, a neural network, a fuzzy rule, a sub-sampling architecture, and the like.
- the basic control unit 110 inputs external world information from various sensors 20 and determines and outputs a basic injection amount and a transient correction rate based on the input external world information.
- the evolution adaptation layer 300 repeatedly performs the autonomous evaluation with the interactive optimization unit 310 that optimizes the control characteristics (response characteristics) of the electronic throttle 12 by GA while repeatedly performing the interactive evaluation. It is composed of an autonomous optimization unit 350 that optimizes the control characteristics (fuel efficiency characteristics) of the engine 10 using GA.
- the interactive optimizing unit 310 has at least one control module that outputs the valve opening of the electronic throttle 12 based on external world information, and performs an evaluation of the response based on a dialog with the user. It is configured to optimize the control module by GA while repeatedly performing type evaluation.
- the autonomous optimization unit 350 performs a correction rate (hereinafter, referred to as a correction rate) for correcting the basic injection amount from the reflective layer 100 and the transient correction rate to values according to the user's desire based on the external world information.
- a correction rate hereinafter, referred to as a correction rate
- the one that corrects the basic injection amount is called the evolution correction rate
- the one that corrects the transient correction rate is called the evolutionary transient correction rate.
- the control module of the autonomous optimization section 350 is fixed to the optimal control module, and the control and control based on the evolution correction rate for correcting the basic injection amount from the reflective layer 100 are performed. While the control is performed by the evolutionary transient correction rate for correcting the transient correction rate from the reflection layer 100, the learning layer 200 learns information on the optimal control module. After the information on the optimal control module is learned by the learning layer 200, the output is returned to “1”, and thereafter, the operation is performed according to the instruction of the user. That is, control by the control module of the autonomous optimization unit 350 is performed only during the evolution simulation and during the learning.
- the learning layer 200 includes a learning unit 210 having two neural networks that can be switched between learning and execution, and the learning unit 210 is controlled by one of the dual networks (for execution). While executing, the other neural network (for learning) learns the relationship between the input and output of the optimal control module from the evolutionary adaptation layer 300.
- the neural network that is performing control is switched to the neural network after learning, and the control is performed by the control module obtained from the learning result in the neural network after learning. Is started, and the neural network that was executing the control starts functioning for learning.
- the neural network in the learning layer 200 is set to output “1” in the initial state. Therefore, in the initial state, the reflective layer 100 and the evolution adaptive layer 300 are set in the initial state. Is performed.
- the execution neural network further includes two neural networks.
- the neural network inputs the throttle opening and the engine speed as external information from various sensors 20 and corrects the basic injection amount from the reflective layer 100 based on the input information (hereinafter referred to as a correction rate (hereinafter referred to as a correction rate).
- the correction rate is referred to as a learning correction rate.
- the other neural network inputs the rate of change of the throttle opening and the engine speed as external information from various sensors 20 and outputs the result. Based on the input information, calculate the transient correction rate from the reflective layer 100.
- a correction rate for correction hereinafter, this correction rate is referred to as a learning transient correction rate
- This configuration is the same for the learning neural network.
- the controller 300 adds the learning correction rate from the learning layer 200 and the evolution correction rate from the evolution adaptive layer 300, and adds the result to the basic injection amount from the reflection layer 100. , And this is used as the first multiplication result.
- the learning transient correction rate from the learning layer 200 and the evolutionary transient correction rate from the evolution adaptive layer 300 are added, and the reflection layer 100
- the transient correction rate from the above is multiplied by the result of the addition, and this is used as the second multiplication result.
- the fuel injection amount is calculated by multiplying the first multiplication result by the second multiplication result. This fuel injection amount is output to the engine 10.
- the interactive optimization unit 310 optimizes the control characteristics of the engine 10 by performing an evolution simulation using the interface unit 320 that performs input and output with the user. It is composed of an evolutionary adaptation unit 340 that turns into a computer, and an evaluation unit 340 that calculates the evaluation value of the individual in the GA.
- the interface section 320 is composed of a display section 322 that displays the evaluation values of the individual during the evolution simulation by GA, and an input section 3224 that inputs the evaluation by the user.
- the evaluation value (responsibility, which will be described in detail later) of each individual for each generation is displayed on the display section 312, and the user can experience the vehicle, such as riding comfort.
- the evaluation of each individual is input to the input unit 314 based on the above.
- the evaluation unit 340 includes a response degree calculation unit 342 that calculates a response degree based on external world information.
- the response degree calculation section 342 inputs the throttle opening and the engine speed as external information, calculates the throttle opening change rate and the engine speed change rate, and calculates the engine speed change.
- the degree of response is calculated by dividing the rate of change by the rate of change in the throttle opening, and the calculated degree of response is output to the evolution adaptation unit 330 as an individual evaluation value in the GA.
- the evolution adaptation unit 33.0 has a control module 332.
- Control module 3 3 2 shows for example, the dynamic characteristics of the two control coefficients SP 1 3 SP 2 and the throttle ⁇ and valve opening, showing the static characteristics of the Surodzutoru opening pulp opening 2
- the relationship between throttle opening and valve opening is defined based on the two control coefficients DR (first order delay element) and AG (imperfect differential element), and the throttle opening is input as external information, and the throttle opening is input.
- the opening degree of the valve is determined based on the degree, and is output to the electronic throttle 12.
- FIG. 5 is a diagram showing a data structure of individual information.
- the control coefficient SP 13 SP 2 showing the static characteristics to the upper side, two control coefficients DR showing the dynamic characteristics, is constructed by applying Ri, respectively allocate the AG to the lower side.
- Ri For example, if one control coefficient is composed of 16 bits of data, the individual information is 64 bits of data in total.
- the initial individual information generated when starting the evolution simulation is determined by random numbers for each individual. At this time, it is preferable to limit the random number generation range to a predetermined range in order to guarantee a certain degree of response. In other words, random numbers are not generated in a range where the response level is clearly worse.
- FIG. 6 is a flowchart showing the processing executed by the evolution adaptation unit 330.
- the GA assigns random initial values to each individual and arranges them in the search space, applies a genetic operation called crossover and mutation for each generation, and selects the growth and selection of individuals according to the evaluation value of the individual. By doing so, a set of individuals of the next generation is obtained. By repeating such generation alternation, the objective is to asymptotically approach the optimal solution.
- Crossover is an operation in which at least two individuals are set as parents, and one or more individual descendants are generated by replacing part of the individual information of the individual parent.
- Mutation is an operation that changes a specific part of individual information of an individual with a predetermined probability, and increases diversity within an individual group. Specifically, this is an operation of inverting a specific bit of the individual information. For example, the individual information of a certain individual is set to “000111”, and a mutation is caused at the third position to obtain the individual information of “001111”. Obtain an individual with
- Selection is an operation to leave a better individual in the population to the next generation according to the evaluation value of the individual.
- each individual is selected with a probability proportional to the evaluation value. For example, in a certain generation, the evaluation values of individuals having individual information of “000000”, “111011”, “110111”, and “010111” were “8”, “4”, “2”, and “2”, respectively. I do. The probability that each individual will be selected is “8/16”, “4/16”, “2 16 ”and“ 2/16 ”. Therefore, on average, in the next generation, the number of individuals with “000000” individual information will increase to two, the number of individuals with “111011” individual information will remain one, and the number of individuals with “110111” individual information will remain. Alternatively, an individual group having the individual information of “010111” is obtained such that any of the individuals remains. However, in the evolution adaptation section 330, the individual is selected by the user.
- the processing executed by the evolution adaptation unit 330 will be described.
- C The processing shown in the flowchart of FIG. 6 is, for example, reading a program stored in ROM in advance and following the read program. The CPU executes.
- step S100 it is determined whether or not an evolution start instruction, which is an instruction to start an evolution simulation, has been input from the input unit 3222, and when it is determined that an evolution start instruction has been input. If (Yes), the process proceeds to step S 102, but if not (No), the process waits at step S 100 until an evolution start instruction is input.
- an evolution start instruction which is an instruction to start an evolution simulation
- step S102 an individual group consisting of a set of a predetermined number (for example, nine) of individuals is virtually generated, and individual information is configured for each individual.
- each individual information the control coefficient showing the static characteristics SP!, 2 two control coefficients DR indicating the SP 2 and dynamics allocates AG, more determined individual information of each individual to a random number.
- the individual information of each individual is stored and managed on a storage device such as a RAM.
- step S108 the process proceeds to step S108, and the response level is obtained from the evaluation unit 340.
- the control module 332 is constructed based on the individual information, and control of the electronic throttle 12 is started by the constructed control module 332, and the obtained response level is the evaluation value and the evaluation value for the individual. I do. The higher this evaluation value is, that is, the more excellent the individual is in the evolutionary simulation by GA, the more it can be positioned.
- step S110 it is determined whether or not the processing from steps S106 to S108 has been completed for all individuals in the population, and the processing is performed for all individuals. If it is determined that the process has been completed (Yes), the process proceeds to step S112.
- step S112 the response level, which is the evaluation value of each individual, is displayed on the display unit 324, and the process proceeds to step S114 to input the user's evaluation from the input unit 322. I do.
- control enters the evaluation mode once.
- the evaluation mode when the user sees the evaluation displayed on the display section 324 and selects an individual having a characteristic to be tested, the control module 332 is activated based on the individual information of the individual selected by the user. It is constructed, temporarily fixed, and controlled by its control module 3 3 2. In this way, the user determines the characteristics of each individual displayed on the display section 324 from the riding comfort and the like actually driving, and evaluates the evaluation value of each individual from the riding comfort.
- step S116 the user ends the evaluation of each individual based on the evaluation of the individual represented on the display unit 3224 and the riding comfort when actually driving.
- control is switched to selection mode, where individuals in the population survive or are selected.
- the survival or selection of individuals can be performed, for example, by switching to the selection mode at the input section 3 2 2, selecting some individuals having the user's favorite characteristics from the population while referring to the display screen, and selecting the selected individuals. And erase the other individuals.
- step S122 in which a mutation process for mutating the individual in the GA is performed, and the process proceeds to step S122, in which the user is satisfied with the input from the input unit 3222. It is determined whether or not characteristics that satisfy the user are obtained. If it is determined that characteristics that the user is satisfied are not obtained (No), the process proceeds to step S124, and the number of generation alternations is equal to or greater than a predetermined number. If it is determined that the number is equal to or more than the predetermined number (Yes), the process proceeds to step S126.
- step S126 the evolution start request for starting the evolution simulation is output to the autonomous optimization unit 350, and the process proceeds to step S128, where the evaluation value is the highest among the population. Individuals are extracted, and a predetermined range based on the responsivity, which is the evaluation value of the extracted individuals, is output to the autonomous optimization unit 350 as a responsivity limit range. Return to processing. On the other hand, when it is determined in step S124 that the number of generation alternations is less than the predetermined number (Yes), the process proceeds to step S104.
- step S122 determines that the characteristics satisfying the user have been obtained (Yes).
- step S110 determines whether the processing from steps S106 to S108 has been completed for all individuals in the population (No). If it is determined in step S110 that the processing from steps S106 to S108 has not been completed for all individuals in the population (No), the process proceeds to step S130. Then, the individual information of the next individual in the individual group is read out, and the flow shifts to step S106.
- FIG. 7 is a block diagram illustrating a configuration of the autonomous optimization unit 350.
- the autonomous optimization unit 350 performs the evolution simulation using the GA to optimize the control characteristics of the engine 10 and the evaluation value of the individual in the GA. It is composed of an evaluator 370 for calculation.
- the evaluation unit 370 includes a fuel efficiency calculation unit 372 that calculates the fuel efficiency of the engine 10 based on the fuel injection amount and the distance pulse, and a responsiveness that calculates the responsiveness based on the throttle opening and the engine speed. It consists of a calculation unit 374.
- the fuel efficiency calculator 372 inputs the fuel injection amount and the distance pulse as external information, calculates the fuel consumption by summing the injection amount at the input interval of the distance pulse input every time the vehicle travels a predetermined distance, and calculates the fuel efficiency.
- the fuel efficiency is output to the evolution adaptation unit 360 as the first evaluation value of the individual in the GA.
- the response degree calculator 374 inputs the throttle opening and the engine speed as external information, calculates the rate of change of the throttle opening and the rate of change of the engine speed, and calculates the rate of change of the engine speed as the throttle.
- the degree of response is calculated by dividing by the rate of change of the degree of opening, and the calculated degree of response Is output to the evolution adaptation unit 360 as the second evaluation value of
- the evolution adaptation unit 360 has a control module 362, and the control module 362 is further configured to include two neural networks.
- the neural network 362a receives the throttle opening and the engine speed as external information from various sensors 20 and outputs an evolutionary correction rate based on the input information.
- the other neural network 36 2 b inputs the rate of change of throttle opening and the engine speed as external information from various sensors 20 and outputs an evolutionary transient correction rate based on the input information. ing.
- FIG. 8 is a diagram showing the structures of the neural networks 36 2 a and 36 2 b and the data structure of the individual information.
- the neural network 3 6 2 a has an input layer for inputting the throttle opening: u , an input layer f i2 for inputting the engine speed, and an input layer fu, an intermediate layer f hl for inputting the output from f i2.
- the input layer fu and the intermediate layer f hl are formed by the synapse of the coupling coefficient k fl
- the input layer f i2 and the intermediate layer f hl are formed by the synapse of the coupling coefficient k f2
- the intermediate layer f hl and the output layer ⁇ 01 Is the synapse of the coupling coefficient k i3
- the input layer fu and the intermediate layer f h2 are the synapse of the coupling coefficient k f4
- the input layer f i2 and the intermediate layer f h2 are the synapse of the coupling coefficient k f5
- output layer f. 2 is connected
- the neural network 3 62 b has an input layer a u for inputting the rate of change of the throttle opening, an input layer a i2 for inputting the engine speed, and an intermediate layer for inputting the output from the input layer a il 3 a i2.
- Input the output of the layers a hl , a h2 and the middle layers a hl , a h2 It consists of five perceptrons with an output layer acl that outputs the evolutionary transient correction rate.
- the input layer a u and the middle layer a hl are formed by the synapse of the coupling coefficient k al
- the input layer a i2 and the middle layer a hl are formed by the synapse of the coupling coefficient k a2 and the middle layer a hl and the output layer a Is the synapse of the coupling coefficient k a3
- the input layer a u and the middle layer a h2 are the synapse of the coupling coefficient ka 4
- the input layer a i2 and the middle layer a h2 are the synapse of the coupling coefficient k a5
- the middle layer a h2 and output layer a. 2 is connected to each other by a synapse having a coupling coefficient k a6 .
- the individual information of the individual in the GA is configured by continuously assigning synaptic coupling coefficients k fl to k f6 to the upper side and synaptic coupling coefficients k al to k a6 to the lower side. For example, if one coupling coefficient is composed of 8 bits of data, the individual information is a total of 96 bits of data.
- the initial individual information generated when starting the evolution simulation is determined by random numbers for each individual. In this case, it is preferable to limit the random number generation range to a predetermined range in order to guarantee a certain degree of response. That is, random numbers are not generated in a range where the response is clearly deteriorated.
- FIG. 9 is a flowchart showing the processing executed by the evolution adaptation unit 3.60.
- a program stored in advance in the ROM is read, and the CPU executes according to the read program.
- step S200 it is determined whether or not an evolution start request has been input from the interactive optimization unit 310, and when it is determined that an evolution start request has been input (Yes), The process proceeds to step S202, but if it is determined that this is not the case (No), the process waits in step S200 until an evolution start request is input.
- the interactive optimization unit 3 1 Input from 0, and proceed to step S204 to virtually generate an individual group consisting of a set of a predetermined number (for example, nine) of individuals, and configure individual information for each individual.
- the individual information is assigned a synaptic coupling coefficient in the neural networks 36 2 a and 36 2 b, and the individual information of each individual is determined by a random number.
- the individual information of each individual is stored and managed on a storage device such as a RAM.
- step S210 the process proceeds to step S210 to acquire the fuel efficiency and the response degree from the evaluation unit 370.
- the control module 36 2 is constructed based on the individual information, the engine 10 is started to be controlled by the constructed control module 36 2, and the fuel efficiency and the response obtained as a result are evaluated values for the individual.
- step S212 it is determined whether or not the processing from steps S20 & to S210 has been completed for all individuals in the population, and the processing is performed for all individuals. If it is determined that has been completed (Yes), the flow shifts to step S214.
- step S214 it is determined whether or not the responsivity, which is the second evaluation value, for each individual belongs to the range of the responsivity input in step S202, and the responsivity is determined.
- the process proceeds to step S216, the individual is culled out, and the process proceeds to step S218.
- step S2128 it is determined whether the processing from steps S208 to S210 has been completed for all individuals in the population, and it has been determined that processing has been completed for all individuals. If (Yes), the process proceeds to step S220, but if not (No), the process proceeds to step S214.
- step S220 a population of individuals whose response degree falls within the restricted range is formed by the processing of steps S214 to S218. Therefore, when a predetermined number (for example, half) or more of the individuals are not selected, a selection process for performing survival or selection of the individuals is further performed so that the total number of the population is equal to or less than half of the original number.
- a predetermined number for example, half
- a selection process for performing survival or selection of the individuals is further performed so that the total number of the population is equal to or less than half of the original number.
- the selection processing for example, in addition to the above-described roulette selection processing, an elite priority selection processing, a lower fitness simple selection processing, or the like can be employed.
- step S222 the process proceeds to step S222 to perform a crossover process for crossing individuals in the GA, and then proceeds to step S224 to perform a mutation process in the GA for sudden mutation of the individual. Then, the flow shifts to step S228, and it is determined whether or not the number of generation alternations is equal to or more than a predetermined number. If it is determined that the number is greater than or equal to the predetermined number (Yes), the flow shifts to step S228.
- step S 2208 an individual with the highest evaluation value is extracted from the individual population, an optimal control module is constructed based on the individual information of the extracted individual, and the control module 362 is fixed to the optimal control module. Then, the process proceeds to step S230, in which the learning layer 200 learns the input / output relationship of the control module 362, and proceeds to step S232 to output the output of the control module 362 to " Set to “1” to end a series of processing and return to the original processing.
- step S212 determines whether the processing from steps S208 to S210 has been completed for all individuals in the population (No). If it is determined in step S212 that the processing from steps S208 to S210 has not been completed for all individuals in the population (No), the process proceeds to step S2334. Then, the individual information of the next individual in the individual group is read out, and the routine goes to Step S208.
- the first generation evolution simulation starts.
- the individual information of the first individual in the individual group is read, and the control module 3332 is set based on the read individual information.
- the control of the electronic throttle 12 is started by the constructed control module 332, and the control by the control module 332 is performed for a while.
- the response level is obtained from the evaluation unit 340 through step S108.
- the evaluation value is obtained for each individual through step S112.
- the response level is displayed on the display section 3 2 4.
- the user selects several individuals having his or her favorite characteristics from the individual group while referring to the evaluation of each individual displayed on the display unit 324.
- the individual selected from the population is left, and the other individuals are deleted, thereby surviving or selecting the individual.
- FIG. 10 is a diagram showing an order for optimizing control characteristics of the rule 12;
- the evolution start request is output to the autonomous optimization unit 350 through steps S126 and S128, and the individual with the highest evaluation value is extracted from the population. Then, a predetermined range based on the responsivity, which is the extracted evaluation value of the individual, is output to the autonomous optimization unit 350 as a responsivity limit range.
- the autonomous optimization unit 350 when an evolution start request is input, through steps S200 to S204, a limited range of responsivity is input, and an individual group consisting of a set of nine individuals is obtained. Generated and individual information is configured for each individual. Here, synaptic coupling coefficients in the neural networks 36 2 a and 36 2 b are assigned to the individual information.
- the first generation of evolutionary simulations will begin.
- the individual information of the first individual in the individual group is read, and the control module is set based on the read individual information.
- the control module 362 is constructed, the control of the engine 10 is started by the constructed control module 362, and the control by the control module 362 is performed for a while.
- step S 2 1 With c
- the same procedure that fuel efficiency and the response degree is obtained from the evaluation unit 3 7 0, to base treatment populations from Step S 2 0 8 until S 2 1 0
- step S 2 14 it is determined whether or not the responsivity, which is the second evaluation value, for each individual is within the limited range of the responsivity input in step S 202. Is determined.
- steps S220 to S224 selection processing, crossover processing, and mutation processing are performed.
- the first generation evolution simulation is completed.
- the evolution simulation is repeatedly performed in the same manner until the number of generation alternations is equal to or more than a predetermined number.
- the fuel efficiency characteristics are autonomously optimized so that the response degree falls within the limited range.
- the point within the limit range and located on the maximum curve (dashed line) of the fuel consumption characteristic and the response characteristic is that point.
- the evolution simulation is completed, through step S228, the population Among them, the individual with the highest evaluation value is extracted, an optimal control module is constructed based on the extracted individual information, and the control module 362 is fixed to the optimal control module.
- step S230 the input / output relationship of the control module 362 is learned by the learning layer 200.
- control is performed using the evolutionary correction rate and the evolutionary transient correction rate for input information such as the actual engine speed obtained by the optimal control module.
- the autonomous optimization unit 350 starts executing the control based on the evolution correction rate and the evolutionary transient correction rate
- the learning neural network of the learning layer 200 changes the input / output relationship of the control module 365 to the learning layer. Learn together with the input / output relationship of the neural network functioning for the execution of 200.
- the output of the autonomous optimization unit 350 is performed by the individual whose evaluation function before that is maximized, and the control law does not change with time.
- the input and output between the autonomous optimization unit 350 and the neural network for execution of the learning layer 200 are averaged with a certain step width, and this is used as the input / output data and the teacher data is used as the data.
- Used for updating sets For example, the average engine speed per second is 500 000 [rpm], the average throttle opening is 20, the average intake air temperature is 28 [V], and the average atmospheric pressure is 101 3 [hP a ] If this is the case, the sum of these and the output of the execution neural network in the autonomous optimization unit 350 and the learning layer 200 at that time is used as the input / output data (see FIG. 11). This input / output data is added to the previous teacher data and a new teacher data set is obtained.
- the old teacher data in the teacher data set whose shortcut distance to the new data is within a certain value is deleted.
- the output is set to “1” for all input data.
- the learning layer 200 learns synapse coupling coefficients in the learning neural network based on the updated teacher data set. The learning of the coupling coefficient is based on the error between the virtual control output obtained from the output of the learning neural network during learning, the basic injection amount from the reflective layer 100, and the transient correction rate, and the actual control output.
- the neural network for learning is used for execution, and the neural network for original control is used for learning.
- the learning layer 200 determines the learning correction rate and the learning transient correction rate by using the newly obtained neural network for execution, and actually outputs them.
- the control module passes through step S2322, The output of 36 2 becomes “1”, and the control by the learning layer 200 and the reflection layer 100 is performed.
- the initial value of the neural network for execution of the learning layer 200 is set so that the output is always "1". By doing so, in the initial state, control can be performed only by the reflection layer 100 and the autonomous optimization unit 350.
- the interactive optimizing unit 310 that controls the response characteristics and the autonomous optimizing unit 350 that controls the fuel consumption characteristics are provided.
- the response calculation unit 342 uses the response calculation unit 342 to output the response when the control coefficient that affects the control characteristics of the interactive optimization unit 310 is input.
- Dialogue degree The optimal solution of the response degree calculation unit 342 is searched by GA while iteratively evaluating by type evaluation, and the autonomous optimization unit 350 is controlled by the autonomous optimization unit 350.
- the optimal solution of the fuel efficiency calculation unit 372 is searched by GA while repeatedly evaluating the fuel efficiency of the fuel efficiency calculation unit 372 by autonomous evaluation so that the response degree falls within the specified range. I have.
- the autonomous optimizing unit 350 can be optimized at relatively high speed without significantly impairing the responsiveness.
- the correction rate of the fuel injection amount of the engine 10 or the correction rate of the transient correction rate is generated by the neural networks 36 2 a and 36 2 b. Assigns synaptic coupling coefficients in the neural networks 36 2 a and 36 2 b ⁇
- the GA is the optimization algorithm described in claims 1, 3, 5 to 7, 7, 8, 10, or 12, or the optimization algorithm described in claim 8.
- the synaptic coupling coefficients in the neural networks 36 2 a and 36 2 b correspond to the control coefficients described in the third, fifth, eighth or tenth claims.
- the RAM corresponds to the storage means described in claims 2 or 4, and the evaluation section 340, 370 stores the evaluation means described in claims 2, 4, or 8.
- the input unit 3222 corresponds to the evaluation input unit described in claims 2 or 4, and the engine 10 and the electronic throttle 12 correspond to the third to fifth claims. , 7, 8, or 10.
- the interactive optimization unit 310 corresponds to the first control system described in Claim 5
- the autonomous optimization unit 350 corresponds to Claim 5.
- the response degree calculation unit 342 corresponds to the first evaluation function described in claim 5 or 7
- the fuel efficiency calculation unit 372 corresponds to the second control system described in claim 7. It corresponds to the second evaluation function described in 5 or 7.
- Steps S118, S122, S222, and S224 correspond to the individual information operating means described in claim 8, and steps S116, S214.
- SS220 corresponds to the individual selecting means described in claim 8.
- the limited range of the responsiveness is set, and while the autonomous evaluation is repeatedly performed, the responsiveness falls within the limited range.
- the fuel efficiency characteristics are optimized as described above, but the invention is not limited to this.
- a limited range is set for the fuel efficiency characteristics, and the interactive evaluation is repeated.
- the response characteristics may be optimized so that the fuel efficiency falls within the limited range.
- the response characteristics may be optimized so that the fuel efficiency falls within the limited range.
- the fuel injection amount is handled as the control output.
- the control output includes, for example, injection time, ignition timing, The intake valve timing, electronic throttle opening, valve lift, exhaust valve timing, or intake / exhaust control valve timing may be considered.
- the intake control valve is a valve provided on the intake pipe for controlling tumble and swirl
- the exhaust control valve is a valve provided on the exhaust pipe for controlling exhaust pulsation. It is.
- the learning layer 200 is configured by a hierarchical neural network, but the configuration of the control system of the learning layer 200 is not limited to the present embodiment.
- CMAC may be used. Advantages of using CMAC include the ability to perform additional learning and faster learning as compared to hierarchical neural networks.
- the correction rate of the fuel injection amount of the engine 10 or the correction rate of the transient correction rate is generated by the neural networks 36 2 a and 36 2 b.
- the information is configured to assign the synaptic coupling coefficients in the neural networks 36 2 a and 36 2 b.
- the present invention is not limited to this, and the individual information may be directly assigned a correction rate of the fuel injection amount of the engine 10 or a correction rate of the transient correction rate. As a result, it is possible to determine the correction rate of the fuel injection amount of the engine 10 or the correction rate of the transient correction rate, which can be expected to obtain a high evaluation value.
- the correction rate of the fuel injection amount or the correction rate of the transient correction rate of the engine 10 is configured to be generated by the neural networks 362a, 362b.
- the fuel injection amount, the transient correction amount, the correction amount of the fuel injection amount, or the correction amount of the transient correction amount may be generated by the neural networks 362a and 362b. This is the same for the configuration in which the calculation is performed directly without being generated by the neural networks 362a and 362b.
- the program may be read into a RAM from a recording medium on which the program indicating the procedure is recorded, and may be executed.
- the recording medium is a semiconductor recording medium such as RAM or ROM, a magnetic recording type recording medium such as FD or HD, an optical reading type recording medium such as CD, CD V, LD, DVD, or a magnetic recording medium such as MO.
- Type / optical reading type recording media including all types of recording media that can be read in a short time, regardless of electronic, magnetic, optical, etc. It is. Industrial applicability
- the control device of the controlled object by the optimization algorithm described in claim 5 after the optimization of the first control system is completed, it is obtained as an optimal evaluation value of the first evaluation function. It is also possible to obtain an effect that the second control system can be optimized at a relatively high speed without significantly impairing the obtained evaluation value.
- the first control system can be optimized at a relatively high speed, and the first control system can be optimized. After the optimization is completed, an effect is obtained that the second control system can be optimized without significantly impairing the evaluation value obtained as the optimum evaluation value of the first evaluation function.
- control device for the control object by the optimization algorithm according to claim 9 of the present invention, it is possible to expect to obtain a high evaluation value, the fuel injection amount, the transient correction amount, and the correction value of the fuel injection amount. Or determine the correction value of the transient correction amount. There is also obtained an effect that it can be set.
- the rotation change characteristic is provided to the user to reduce power consumption. There is also obtained an effect that characteristics can be optimized according to predetermined evaluation criteria.
Landscapes
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Automation & Control Theory (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Feedback Control In General (AREA)
- Combined Controls Of Internal Combustion Engines (AREA)
- Electrical Control Of Air Or Fuel Supplied To Internal-Combustion Engine (AREA)
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US10/467,872 US7062333B2 (en) | 2001-02-23 | 2002-02-05 | Optimal solution search device, device for controlling controlled object by optimizing algorithm, and optimal solution search program |
EP02710510A EP1372107A4 (en) | 2001-02-23 | 2002-02-05 | SEARCH FOR OPTIMAL SOLUTIONS, DEVICE FOR CONTROLLING A CONTROLLED OBJECT BY OPTIMIZING ALGORITHM AND SEARCH PROGRAM FOR OPTIMUM SOLUTIONS |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2001049342A JP2002251597A (ja) | 2001-02-23 | 2001-02-23 | 最適解探索装置、最適化アルゴリズムによる制御対象の制御装置及び最適解探索プログラム |
JP2001-49342 | 2001-02-23 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2002069257A1 true WO2002069257A1 (fr) | 2002-09-06 |
Family
ID=18910457
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2002/000932 WO2002069257A1 (fr) | 2001-02-23 | 2002-02-05 | Dispositif de recherche de solution optimale, dispositif pour commander un objet commande par algorithme d'optimisation, et programme de recherche de solution optimale |
Country Status (4)
Country | Link |
---|---|
US (1) | US7062333B2 (ja) |
EP (1) | EP1372107A4 (ja) |
JP (1) | JP2002251597A (ja) |
WO (1) | WO2002069257A1 (ja) |
Families Citing this family (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002251597A (ja) * | 2001-02-23 | 2002-09-06 | Yamaha Motor Co Ltd | 最適解探索装置、最適化アルゴリズムによる制御対象の制御装置及び最適解探索プログラム |
US7035834B2 (en) * | 2002-05-15 | 2006-04-25 | Caterpillar Inc. | Engine control system using a cascaded neural network |
DE10354322B4 (de) * | 2003-11-20 | 2022-06-09 | Bayerische Motoren Werke Aktiengesellschaft | Verfahren und System zur Ermittlung der Fahrsituation |
JP2005271824A (ja) * | 2004-03-25 | 2005-10-06 | Mitsubishi Fuso Truck & Bus Corp | 車両の挙動制御装置 |
US7835786B2 (en) * | 2005-07-25 | 2010-11-16 | Wisconsin Alumni Research Foundation | Methods, systems, and computer program products for optimization of probes for spectroscopic measurement in turbid media |
US7291934B2 (en) * | 2005-08-30 | 2007-11-06 | Caterpillar Inc. | Machine with an electrical system |
JP4561652B2 (ja) * | 2006-03-01 | 2010-10-13 | 株式会社デンソー | 内燃機関の制御装置 |
WO2007109126A2 (en) * | 2006-03-17 | 2007-09-27 | Duke University | Monte carlo based model of fluorescence |
US7751039B2 (en) * | 2006-03-30 | 2010-07-06 | Duke University | Optical assay system for intraoperative assessment of tumor margins |
US20080270091A1 (en) * | 2007-02-23 | 2008-10-30 | Nirmala Ramanujam | Scaling method for fast monte carlo simulation of diffuse reflectance spectra from multi-layered turbid media and methods and systems for using same to determine optical properties of multi-layered turbid medium from measured diffuse reflectance |
KR101399199B1 (ko) * | 2007-07-16 | 2014-05-27 | 삼성전자주식회사 | 소프트웨어 로봇의 유전자 코드 구성 방법 |
WO2009043050A2 (en) * | 2007-09-27 | 2009-04-02 | Duke University | Optical assay system with a multi-probe imaging array |
US9820655B2 (en) * | 2007-09-28 | 2017-11-21 | Duke University | Systems and methods for spectral analysis of a tissue mass using an instrument, an optical probe, and a Monte Carlo or a diffusion algorithm |
WO2010042249A2 (en) * | 2008-04-24 | 2010-04-15 | Duke University | A diffuse reflectance spectroscopy device for quantifying tissue absorption and scattering |
US20100049561A1 (en) * | 2008-08-22 | 2010-02-25 | Alstom Technology Ltd. | Fluidized bed combustion optimization tool and method thereof |
US8483949B2 (en) | 2009-04-13 | 2013-07-09 | Toyota Jidosha Kabushiki Kaisha | Running pattern calculating apparatus and running pattern calculating method |
JP4821879B2 (ja) * | 2009-04-13 | 2011-11-24 | トヨタ自動車株式会社 | 走行軌跡演算装置、および、走行軌跡演算方法 |
DE112009005242B4 (de) * | 2009-09-18 | 2015-02-12 | Honda Motor Co., Ltd. | Regelungs-/Steuerungssystem für einen Verbrennungsmotor |
DE112009005254B4 (de) * | 2009-09-18 | 2015-11-05 | Honda Motor Co., Ltd. | Regelungs-/Steuerungssystem für einen Verbrennungsmotor |
JP6857332B2 (ja) * | 2018-03-13 | 2021-04-14 | オムロン株式会社 | 演算装置、演算方法、及びそのプログラム |
JP6702380B2 (ja) * | 2018-09-14 | 2020-06-03 | トヨタ自動車株式会社 | 内燃機関の制御装置 |
JP6593560B1 (ja) * | 2019-02-15 | 2019-10-23 | トヨタ自動車株式会社 | 内燃機関の失火検出装置、内燃機関の失火検出システム、データ解析装置、および内燃機関の制御装置 |
US10947919B1 (en) | 2019-08-26 | 2021-03-16 | Caterpillar Inc. | Fuel injection control using a neural network |
US11603111B2 (en) * | 2019-10-18 | 2023-03-14 | Toyota Jidosha Kabushiki Kaisha | Vehicle controller, vehicle control system, and learning device for vehicle |
JP7205503B2 (ja) | 2020-01-22 | 2023-01-17 | トヨタ自動車株式会社 | 内燃機関の制御装置 |
JP7222366B2 (ja) * | 2020-01-27 | 2023-02-15 | トヨタ自動車株式会社 | 内燃機関の制御装置 |
JP7359011B2 (ja) | 2020-02-05 | 2023-10-11 | トヨタ自動車株式会社 | 内燃機関の制御装置 |
CN115166449B (zh) * | 2022-08-11 | 2024-07-23 | 云南电网有限责任公司电力科学研究院 | 氧化锌阀片性能评估方法和系统 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6032139A (en) * | 1996-09-27 | 2000-02-29 | Yamaha Hatsudoki Kabushiki Kaisha | Electronic controller using genetic evolution techniques suitable for controlling a motor |
EP1033637A2 (en) * | 1999-03-02 | 2000-09-06 | Yamaha Hatsudoki Kabushiki Kaisha | Method and apparatus for optimizing overall characteristic of device, using heuristic method |
EP1039356A1 (en) * | 1999-03-24 | 2000-09-27 | Yamaha Hatsudoki Kabushiki Kaisha | Overall characteristic optimization method and apparatus therefor |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5418858A (en) * | 1994-07-11 | 1995-05-23 | Cooper Tire & Rubber Company | Method and apparatus for intelligent active and semi-active vibration control |
US5877954A (en) * | 1996-05-03 | 1999-03-02 | Aspen Technology, Inc. | Hybrid linear-neural network process control |
US5933345A (en) * | 1996-05-06 | 1999-08-03 | Pavilion Technologies, Inc. | Method and apparatus for dynamic and steady state modeling over a desired path between two end points |
US6381504B1 (en) * | 1996-05-06 | 2002-04-30 | Pavilion Technologies, Inc. | Method for optimizing a plant with multiple inputs |
JP3825845B2 (ja) * | 1996-09-27 | 2006-09-27 | ヤマハ発動機株式会社 | 進化的制御方式 |
US5963458A (en) * | 1997-07-29 | 1999-10-05 | Siemens Building Technologies, Inc. | Digital controller for a cooling and heating plant having near-optimal global set point control strategy |
JP2000020103A (ja) * | 1998-07-02 | 2000-01-21 | Yamaha Motor Co Ltd | 遺伝的アルゴリズムの評価方法 |
JP2002251597A (ja) * | 2001-02-23 | 2002-09-06 | Yamaha Motor Co Ltd | 最適解探索装置、最適化アルゴリズムによる制御対象の制御装置及び最適解探索プログラム |
-
2001
- 2001-02-23 JP JP2001049342A patent/JP2002251597A/ja active Pending
-
2002
- 2002-02-05 EP EP02710510A patent/EP1372107A4/en not_active Withdrawn
- 2002-02-05 WO PCT/JP2002/000932 patent/WO2002069257A1/ja not_active Application Discontinuation
- 2002-02-05 US US10/467,872 patent/US7062333B2/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6032139A (en) * | 1996-09-27 | 2000-02-29 | Yamaha Hatsudoki Kabushiki Kaisha | Electronic controller using genetic evolution techniques suitable for controlling a motor |
EP1033637A2 (en) * | 1999-03-02 | 2000-09-06 | Yamaha Hatsudoki Kabushiki Kaisha | Method and apparatus for optimizing overall characteristic of device, using heuristic method |
EP1039356A1 (en) * | 1999-03-24 | 2000-09-27 | Yamaha Hatsudoki Kabushiki Kaisha | Overall characteristic optimization method and apparatus therefor |
Non-Patent Citations (1)
Title |
---|
See also references of EP1372107A4 * |
Also Published As
Publication number | Publication date |
---|---|
EP1372107A4 (en) | 2006-03-08 |
US7062333B2 (en) | 2006-06-13 |
JP2002251597A (ja) | 2002-09-06 |
US20040078095A1 (en) | 2004-04-22 |
EP1372107A1 (en) | 2003-12-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2002069257A1 (fr) | Dispositif de recherche de solution optimale, dispositif pour commander un objet commande par algorithme d'optimisation, et programme de recherche de solution optimale | |
JP3825845B2 (ja) | 進化的制御方式 | |
Stanley et al. | Efficient reinforcement learning through evolving neural network topologies | |
Thiele et al. | A preference-based evolutionary algorithm for multi-objective optimization | |
JP5124230B2 (ja) | パラメトリック多目的最適化装置、パラメトリック多目的最適化方法およびパラメトリック多目的最適化プログラム | |
US6721647B1 (en) | Method for evaluation of a genetic algorithm | |
Ochoa | Error thresholds in genetic algorithms | |
CN107197006B (zh) | 基于全局QoS分解的多约束服务选取方法及其装置 | |
US6324529B1 (en) | Evolutionary controlling system | |
US6529816B1 (en) | Evolutionary controlling system for motor | |
JP2009099051A (ja) | パラメトリック多目的最適化装置、パラメトリック多目的最適化方法およびパラメトリック多目的最適化プログラム | |
CN113830097B (zh) | 车辆、模型学习系统以及服务器 | |
CN113138555A (zh) | 一种基于遗传算法优化的grnn电主轴热误差建模方法 | |
Fiscko et al. | Efficient solutions for targeted control of multi-agent mdps | |
Chen et al. | Generative adversarial reward learning for generalized behavior tendency inference | |
CN112528556B (zh) | 一种基于集成模型辅助社会学习粒子群算法的微机电系统设计优化方法 | |
Chidambaran et al. | Multi-criteria evolution of neural network topologies: Balancing experience and performance in autonomous systems | |
JP2002245434A (ja) | 進化的手法による最適解探索装置、進化的手法による制御対象の制御装置及び進化的手法による最適解探索プログラム | |
JP2002251598A (ja) | 最適解探索装置、制御対象の制御装置及び最適解探索プログラム | |
JP2002251599A (ja) | 進化的手法による最適解探索装置、進化的手法による制御対象の制御装置及び進化的手法による最適解探索プログラム | |
JP2000339005A (ja) | 制御対象の最適化制御方法及び制御装置 | |
CN115409661A (zh) | 基于集成学习模型的就业推荐方法、系统、设备及介质 | |
JPH10154003A (ja) | 進化的制御方式 | |
Júnior et al. | Hybrid metaheuristics using reinforcement learning applied to salesman traveling problem | |
JPH10333705A (ja) | 総合制御方式 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): US |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR |
|
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 10467872 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2002710510 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 2002710510 Country of ref document: EP |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 2002710510 Country of ref document: EP |