CONTROL SYSTEM MANUFACTURING APPARATUS, CONTROL SYSTEM MANUFACTURING METHOD, AND PROGRAM
The control system production apparatus and method address the challenge of simulating nonlinear characteristics by generating a first and second model, enhancing simulation accuracy and control precision.
Patent Information
- Application Number
- JP2022555003
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-10-06
- Publication Date
- 2024-11-06
- Estimated Expiration
- 2040-10-06
AI Technical Summary
Existing control system production methods lack accuracy in simulation due to the difficulty in modeling nonlinear characteristics and components such as backlash, Coulomb friction, and quantization errors in controlled objects.
A control system production apparatus and method that generate a first model based on a command and response profile, and a second model representing characteristics not included in the first model, using a combination of linear and nonlinear unit models, to improve simulation accuracy.
Enhances simulation accuracy by effectively modeling nonlinear characteristics and improving the relationship between commands and control amounts, leading to more precise control of controlled objects.
Smart Images

Figure 0007578244000001 
Figure 0007578244000002 
Figure 0007578244000003
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a control system production device, a control system production method, and a program. [Background technology]
[0002] Patent Document 1 discloses a robot simulator that includes a model storage unit that stores model information regarding a robot and obstacles, and an information processing unit that generates, based on the model information, a path along which the tip of the robot can move from a start position to an end position while avoiding collisions between the robot and obstacles. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2018-134703 A Summary of the Invention [Problem to be solved by the invention]
[0004] The present disclosure provides a control system production device and a control system production method that are effective in improving the accuracy of a simulation. [Means for solving the problem]
[0005] A production apparatus for a control system according to one aspect of the present disclosure includes a first model generation unit that generates a first model representing at least a part of a response characteristic of the control object to a command based on a command profile that represents a time change of a command for the control object and a response profile that represents a time change of a state of the control object according to the command profile, a second model generation unit that generates a second model that represents a characteristic of the response characteristics not included in the first model based on the command profile, the response profile, and the first model, and a parameter generation unit that generates control parameters that define a relationship between the command and a control amount for causing the control object to follow the command, based on the first model and the second model.
[0006] A method for producing a control system according to another aspect of the present disclosure includes generating a first model representing at least a portion of the response characteristics of the controlled object to a command based on a command profile representing a time change of a command for the controlled object and a response profile representing a time change of a state of the controlled object according to the command profile; generating a second model representing characteristics of the response characteristics not included in the first model based on the command profile, the response profile, and the first model; and generating control parameters defining a relationship between the command and a control amount for causing the controlled object to follow the command, based on the first model and the second model.
[0007] A program according to yet another aspect of the present disclosure causes an apparatus to execute the following operations: generating a first model representing at least a portion of the response characteristics of the control object to a command, based on a command profile representing a time change of a command to the control object and a response profile representing a time change of a state of the control object according to the command profile; generating a second model representing characteristics of the response characteristics not included in the first model, based on the command profile, the response profile, and the first model; and generating control parameters, based on the first model and the second model, that define the relationship between the command and a control amount for causing the control object to follow the command. Effect of the Invention
[0008] According to the present disclosure, it is possible to provide a control system production device and a control system production method that are effective in improving the accuracy of a simulation. [Brief description of the drawings]
[0009] [Figure 1] FIG. 2 is a schematic diagram illustrating a configuration of a control system. [Diagram 2] FIG. 2 is a schematic diagram illustrating a hardware configuration of a control device. [Diagram 3] 4 is a flowchart illustrating a control procedure of a motor. [Figure 4]1 is a flowchart illustrating a parameter generation procedure. [Diagram 5] 10 is a flowchart illustrating a control procedure for a motor based on a control parameter. [Figure 6] FIG. 1 is a schematic diagram illustrating a schematic configuration of a robot. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] Hereinafter, the embodiments will be described in detail with reference to the drawings. In the description, the same elements or elements having the same functions are denoted by the same reference numerals, and duplicated description will be omitted.
[0011] [Control System] As shown in Fig. 1, a control system 1 according to this embodiment is a system including a control target 2 and a control device 100 that controls the control target 2. The control target 2 includes a motor 3, a transmission mechanism 5, a driven object 4, and a sensor 6. There are no particular limitations on the shape, structure, and size of the driven object 4 in the control target 2, and on the number and size of the motors 3. In the following, a case will be illustrated in which there is one motor 3 and the driven object 4 is one rigid body.
[0012] The motor 3 generates power for driving the object to be driven 4 in response to the supply of electric power. The motor 3 may be a rotary motor that rotates the object to be driven 4, or a linear motor that displaces the object to be driven 4 along a straight line. The motor 3 may be a synchronous motor or an induction motor. The motor 3 may be a permanent magnet type synchronous motor such as a surface permanent magnet (SPM) motor or an interior permanent magnet (IPM) motor, or may be a synchronous motor without a permanent magnet such as a synchronous reluctance motor.
[0013] The transmission mechanism 5 transmits the speed of the output shaft of the motor 3 to the driven object 4. The transmission mechanism 5 is, for example, a reducer configured to reduce the speed of the driven object 4 relative to the speed of the output shaft of the motor 3. For example, the transmission mechanism 5 includes a plurality of gears that transmit power from the motor 3 to the driven object 4 at a predetermined reduction ratio.
[0014] The sensor 6 detects a response value of the controlled object 2 corresponding to the command. A specific example of the response value is the position or speed of the driven object 4. If the motor 3 is a rotary type, the rotation angle of the driven object 4 by the motor 3 corresponds to the "position", and the rotation speed of the driven object 4 by the motor 3 corresponds to the "speed". Furthermore, torque corresponds to the driving force. A specific example of the sensor 6 is a rotary encoder that outputs a pulse signal with a frequency proportional to the operating speed of the driven object 4. With a rotary encoder, both the position and the speed of the driven object 4 can be obtained.
[0015] The control device 100 controls the motor 3 so as to cause the driven object 4 to perform a desired operation. The control device 100 controls the motor 3 based on control parameters generated in advance. The control parameters determine the relationship between an operation command for the controlled object 2 and at least a part of a control amount for making the controlled object 2 follow the operation command. Specific examples of the operation command include a position command and a speed command for the driven object 4. Specific examples of the control amount include a drive force command for the motor 3.
[0016] Specific examples of the control parameters include proportional gain, integral gain, differential gain, etc. in so-called feedback control. Another example of the control parameters is a feedforward compensation parameter that compensates for a response delay in feedback control.
[0017] These control parameters can be generated based on a model of the control object 2. The model of the control object 2 represents at least a part of the response characteristics of the control object 2 to a control command. The control device 100 may be configured to generate the model of the control object 2 and generate the control parameters based on the model.
[0018] Generating the model and the control parameters corresponds to producing at least a part of the control device 100. Therefore, it can be said that the control device 100 having the function of generating the model and the control parameters includes a production device of the control device 100 itself (a production device of the control system 1). The production device does not necessarily have to be included in the control device 100, and may be included in another device capable of communicating with the control device 100. Specific examples of the other device include a higher-level controller (e.g., a programmable logic controller), a personal computer, or a mobile terminal capable of communicating with the control device 100.
[0019] Here, in order to generate appropriate control parameters, it is necessary to generate an appropriate model. A specific example of the model is a linear model represented by a linear differential equation. However, the actual control target 2 contains components that are difficult to model using a linear model. Specific examples of components that are difficult to model include rattling due to backlash in the transmission mechanism 5, Coulomb friction occurring inside the control target 2, and quantization errors in digital signal processing. The control device 100 generates a model that includes components that are difficult to model using a linear model, and generates control parameters based on the generated model.
[0020] For example, the control device 100 is configured to generate a first model representing at least a part of the response characteristics of the control object 2 based on a command profile representing the time change of a command for the control object 2 and a response profile representing the time change of the state of the control object 2 according to the command profile; generate a second model representing the response characteristics that are not included in the first model based on the command profile, the response profile, and the first model; and generate control parameters based on the first model and the second model.
[0021] A specific example of the command is a command for the driving force to be generated by the motor 3 (hereinafter referred to as a "driving force command"), but is not limited to this. The command may be a command for the speed of the motor 3 or a command for the position of the motor 3.
[0022] Specific examples of the response include, but are not limited to, the speed of the driven object 4 and the position of the driven object 4. The response may be a driving force transmitted from the motor 3 to the driven object 4, or a driving current supplied to the motor 3 to transmit the driving force. Below, a specific example will be given of a case where the command is a driving force command and the response is the speed or position of the driven object 4.
[0023] For example, the control device 100 includes, as functional components (hereinafter referred to as "functional blocks"), a command profile memory unit 111, a control unit 112, a response profile memory unit 113, a first model generation unit 114, a gain profile memory unit 115, a second model generation unit 132, a model memory unit 121, a parameter generation unit 141, and a parameter memory unit 142.
[0024] The command profile storage unit 111 stores the command profile. The command profile may include vibrations that repeatedly increase and decrease over time. Examples of the command profile include various signals such as a chirp signal, a multi-sine signal, a pseudo-random signal, a random signal, a short-time irregular signal, and an impulse vibration signal. The command profile storage unit 111 may store a command profile that is continuously represented as a function, or may store a command profile that is discretely represented as a time-series command value group (for example, a table).
[0025] The control unit 112 drives the motor 3. For example, the control unit 112 generates driving power based on the command profile stored in the command profile storage unit 111, and supplies the generated driving power to the motor 3, repeatedly at a predetermined control period.
[0026] When driving the motor 3, the control unit 112 obtains a response profile that indicates a time change in the state of the control target 2 (e.g., the speed and position of the drive target 4) corresponding to the command profile, and stores the response profile in the response profile storage unit 113. For example, the control unit 112 obtains the detection value of the state of the drive target 4 by the sensor 6 in each control cycle during the period in which the control unit 112 drives the motor 3 based on the command profile, and stores the detection value in the response profile storage unit 113 in chronological order.
[0027] The first model generating unit 114 generates the first model based on the command profile stored in the command profile storage unit 111 and the response profile stored in the response profile storage unit 113 in correspondence with the command profile. For example, the first model generating unit 114 generates a gain profile representing the relationship between the frequency of the command and the response gain of the control target 2 to the command based on the command profile and the response profile, and stores the gain profile in the gain profile storage unit 115. For example, the first model generating unit 114 generates the gain profile by a frequency analysis method such as FFT (Fast Fourier Transform). The response gain is a value representing the relationship between the amplitude of the response profile and the amplitude of the command profile, and is, for example, a magnification of the amplitude of the response profile to the amplitude of the command profile.
[0028] The first model generating unit 114 generates the first model based on a linear model in which the frequency characteristics are fitted to the gain profile. The linear model is a model in which the relationship between an input function that represents the relationship between elapsed time and an input, and an output function that represents the relationship between elapsed time and an output, is expressed by a linear differential equation. "Fitting" does not mean to make them completely match, but to make them approximate. Specific examples of fitting methods include the least squares method, polynomial approximation, circular approximation, and approximation using various functions (trigonometric functions, B-Spline, etc.).
[0029] The first model generating unit 114 may generate the first model based on a trained model generated by machine learning so as to represent the relationship between the gain profile and the linear model. The first model generating unit 114 stores the generated first model in the model storage unit 121.
[0030] The second model generation unit 132 generates the second model based on the command profile, the response profile, and the first model. The second model generation unit 132 may generate the second model based on a plurality of types of unit models prepared in advance. For example, the control device 100 further includes a unit model storage unit 131 that stores the plurality of types of unit models, and the second model generation unit 132 generates the second model based on the plurality of types of unit models stored in the unit model storage unit 131.
[0031] At least one of the multiple types of unit models may be a nonlinear model that represents a nonlinear characteristic with respect to a command. A specific example of the unit model is a model that represents the relationship between the speed of the motor 3 and the Coulomb friction acting on the motor 3. As the speed switches from a negative value to a positive value, the Coulomb friction also changes in a stepwise manner from a negative value to a positive value, so the relationship between the speed and the Coulomb friction is nonlinear.
[0032] Another example of the unit model is a model that expresses the relationship between the difference between the position of the motor 3 and the position of the driven object 4, and the magnitude of the torque transmitted from the motor 3 to the driven object 4. The difference between the position of the motor 3 and the position of the driven object 4 is caused by backlash of the transmission mechanism 5 described above. Due to the presence of backlash, the torque transmitted from the motor 3 to the driven object 4 changes stepwise between a state in which the gears are not meshed and a state in which the gears are meshed, so the relationship between the difference between the position of the motor 3 and the position of the driven object 4 and the torque transmitted from the motor 3 to the driven object 4 is nonlinear. Another example of the unit model is a model that expresses a quantization error when the state of the controlled object 2 is digitized by the sensor 6. The second model generation unit 132 stores the generated second model in the model storage unit 121.
[0033] The second model generation unit 132 may generate the second model tentatively based on a weighted sum of multiple types of unit models, and change the weights of the multiple types of unit models in the second model based on an evaluation result of the modeling error between the first model and the second model, repeating this process until a predetermined condition is satisfied.
[0034] For example, the control device 100 further includes a model evaluation unit 122 and a virtual profile storage unit 123. The model evaluation unit 122 generates a virtual response profile by simulating a time change in a state of the control target 2 according to the command profile based on the command profile stored in the command profile storage unit 111 and the first model and the second model stored in the model storage unit 121, and stores the virtual response profile in the virtual profile storage unit 123.
[0035] The model evaluation unit 122 evaluates an error of the hypothetical response profile with respect to the response profile, based on the response profile stored in the response profile storage unit 113 and the hypothetical response profile stored in the hypothetical profile storage unit 123. The error of the hypothetical response profile with respect to the response profile is an example of the above-mentioned modeling error.
[0036] The model evaluation unit 122 is not particularly limited in the calculation method for evaluating the error as long as it can output a numerical value representing the evaluation result of the error. For example, the model evaluation unit 122 may output a value obtained by integrating the absolute value of the error over time as the evaluation result, or may output a time average value of the absolute value of the error as the evaluation result. In addition, the model evaluation unit 122 may output a value obtained by integrating the square of the error over time as the evaluation result, or may output a time average value of the square of the error as the evaluation result, or may output a square root of the time average value as the evaluation result. The model evaluation unit 122 may output the evaluation result of the error every time the contents stored in the model storage unit 121 are updated.
[0037] The second model generation unit 132 provisionally generates a second model based on the weighted sum of multiple types of unit models, and stores it in the model storage unit 121. When the stored contents of the model storage unit 121 are updated, the model evaluation unit 122 evaluates the error. The second model generation unit 132 changes the weights of multiple types of unit models in the second model based on the error evaluation result by the model evaluation unit 122, and stores the changed second model in the model storage unit 121, repeating this process until a predetermined condition is satisfied. When the predetermined condition is satisfied, the second model stored in the model storage unit 121 becomes the generation result of the second model by the second model generation unit 132.
[0038] A specific example of the predetermined condition is that the error evaluation result is equal to or lower than a predetermined level (hereinafter referred to as the "second level"). In this case, the second model generation unit 132 may store the second model that satisfies the predetermined condition in the model storage unit 121 as the actual generation result. An example of the predetermined condition is that the number of times the weights of multiple types of unit models in the second model are changed reaches a predetermined number. In this case, the second model generation unit 132 may store the second model that has the smallest error evaluation result among the second models generated until the predetermined condition is satisfied in the model storage unit 121 as the actual generation result.
[0039] The second model generating unit 132 may generate an error model that represents a relationship between weights of a plurality of types of unit models in the second model and an error evaluation result based on the provisionally generated second model and an error evaluation result based on the provisionally generated second model, change the weights based on a predetermined target error and the error model, and update the error model based on the second model after the weights have been changed and an error evaluation result based on the second model after the weights have been changed, repeating this process until a predetermined condition is satisfied to generate the second model. A specific example of the error model is a statistical model based on Bayesian estimation. The predetermined target error is equal to or less than the second level.
[0040] For example, the control device 100 further includes a learning data storage unit 133 and an error model storage unit 134. The second model generation unit 132 stores in the learning data storage unit 133 a record in which the provisionally generated second model and an evaluation result of an error based on the provisionally generated second model are associated with each other. The second model generation unit 132 generates an initial error model based on the record stored in the learning data storage unit 133, and stores the generated initial error model in the error model storage unit 134. The second model generation unit 132 may randomly change the weight and store the record corresponding to the second model after the weight change in the learning data storage unit 133 one or more times, thereby generating an initial error model based on a plurality of records accumulated in the learning data storage unit 133.
[0041] After generating the initial error model, the second model generation unit 132 repeats the following procedure until a predetermined condition is satisfied. i) The weights are changed based on the target error and the error model, and the second model after the weights are changed is stored in the model storage unit 121. ii) A record associating the second model after the weighting change with the evaluation result of the error based on the second model after the weighting change is stored in the learning data storage unit 133. iii) The error model in the error model storage unit 134 is updated based on the records stored in the learning data storage unit 133.
[0042] The first model generation unit 114 may regenerate the first model based on the command profile, the response profile, and the second model when the error evaluation result by the model evaluation unit 122 after the second model is generated exceeds a predetermined level. The predetermined level (hereinafter, referred to as the "first level") here may be a value lower than the above-mentioned second level (the second level may be a value higher than the first level).
[0043] For example, when the error evaluation result after the second model is generated exceeds the first level, the first model generation unit 114 may change the gain profile based on the command profile, the response profile, and the second model, and fit the frequency characteristics of the linear model to the changed gain profile to regenerate the first model.
[0044] For example, the first model generation unit 114 may generate a modified profile by removing the response of the second model from the response profile based on the command profile, the response profile, and the second model, and change the gain profile based on the command profile and the modified profile.
[0045] The second model generation unit 132 may regenerate the second model based on the command profile, the response profile, and the first model when the error evaluation result after the first model is regenerated exceeds a predetermined level (for example, the above-mentioned first level). Note that, when the first model is reproduced, the second model generation unit 132 may regenerate the second model without determining whether or not the second model needs to be regenerated.
[0046] In this manner, by regenerating the first model as necessary after the second model is generated, and regenerating the second model as necessary after the first model is regenerated, the results of generating the first model and the results of generating the second model influence each other, thereby further improving the modeling accuracy of the control object 2.
[0047] The parameter generating unit 141 generates a control parameter that defines the relationship between a command and a control amount for making the controlled object 2 follow the command, based on the first model and the second model stored in the model storage unit 121. As an example, the parameter generating unit 141 generates a control parameter that defines the relationship between a command and a feedforward compensation value for making the controlled object 2 follow the command. A specific example of the feedforward compensation value is a compensation value that is added to the control amount by feedback control so as to reduce a control deviation by feedback control (for example, a deviation between a command value and a response value of the controlled object 2 to the command value). The parameter generating unit 141 stores the generated control parameter in the parameter storage unit 142.
[0048] After the control parameters are generated, the control unit 112 controls the motor 3 based on the control parameters. As an example, when controlling the position of the motor 3, the control unit 112 acquires information on the current position and current speed of the driven object 4 based on the detection value of the sensor 6, and generates driving power to make these follow the target position and target speed. As an example, the control unit 112 performs a proportional operation, a proportional-integral operation, or a proportional-integral-differential operation on the deviation between the target position and the current position (hereinafter referred to as the "position deviation") to calculate a target speed for reducing the position deviation. In addition, the control unit 112 performs a proportional operation, a proportional-integral operation, or a proportional-integral-differential operation on the deviation between the target speed and the current speed (hereinafter referred to as the "speed deviation") to calculate a driving force command for reducing the speed deviation.
[0049] Furthermore, the control unit 112 calculates the feedforward compensation value based on at least one of the target position and the target speed and the control parameters stored in the parameter storage unit 142. The control unit 112 adds the feedforward compensation value to the driving force command to calculate a corrected command, generates driving power based on the corrected command, and supplies it to the motor 3. The control unit 112 repeats the above procedure at a predetermined control period.
[0050] FIG. 2 is a block diagram illustrating a hardware configuration of the control device 100. As shown in FIG. 2, the control device 100 has a circuit 190. The circuit 190 includes one or more processors 191, a memory 192, a storage 193, an input / output port 194, and a driver circuit 195. The storage 193 has a computer-readable storage medium, such as a non-volatile semiconductor memory. The storage 193 stores a program that causes the control device 100 to execute the following operations: generating a first model that represents at least a part of the response characteristics of the control target 2 to a command based on a command profile that represents a time change of a command for the control target 2 and a response profile that represents a time change of the state of the control target 2 according to the command profile; generating a second model that represents a characteristic not included in the first model among the response characteristics based on the command profile, the response profile, and the first model; and generating a control parameter that determines a relationship between the command and a control amount for making the control target 2 follow the command based on the first model and the second model.
[0051] The memory 192 temporarily stores the programs loaded from the storage medium of the storage 193 and the results of calculations by the processor 191. The processor 191 configures each functional block of the control device 100 by executing the above programs in cooperation with the memory 192. The input / output port 194 inputs and outputs electric signals to and from the sensor 6 in accordance with instructions from the processor 191. The driver circuit 195 outputs driving power to the motor 3 in accordance with instructions from the processor 191.
[0052] The control device 100 is not necessarily limited to a device that configures each function by a program. For example, the control device 100 may configure at least some of its functions by a dedicated logic circuit or an ASIC (Application Specific Integrated Circuit) that integrates such a dedicated logic circuit.
[0053] [Control parameter generation procedure] Next, as an example of a method for producing a control device, a procedure for generating control parameters executed by the control device 100 will be illustrated. This procedure includes generating a first model representing at least a part of the response characteristics of the control target 2 to a command based on a command profile representing a time change of a command for the control target 2 and a response profile representing a time change of the state of the control target 2 according to the command profile, generating a second model representing a characteristic of the response characteristics not included in the first model based on the command profile, the response profile, and the first model, and generating control parameters that determine a relationship between a command and a control amount for making the control target 2 follow the command based on the first model and the second model. Hereinafter, a procedure for acquiring the response profile and a procedure for generating the control parameters based on the acquired response profile will be described in detail.
[0054] (Procedure for obtaining response profile) As shown in Fig. 3, the control device 100 executes steps S01, S02, and S03 in order. In step S01, the control unit 112 drives the motor 3. For example, the control unit 112 generates driving power based on the command profile stored in the command profile storage unit 111 and supplies it to the motor 3. In step S02, the control unit 112 acquires a detection value of the state of the driven object 4 by the sensor 6, and stores it in chronological order in the response profile storage unit 113. In step S03, the control unit 112 checks whether the supply of driving power based on the command profile has been completed over the entire period of the command profile.
[0055] If it is determined in step S03 that the supply of drive power based on the command profile is not completed for a certain period, the control device 100 returns the process to step S01. After that, the control device 100 repeats the supply of drive power based on the command profile to the motor 3 and the acquisition of the detection value of the state of the driven object 4 by the sensor 6 at a predetermined control period until the supply of drive power based on the command profile is completed for the entire period of the command profile. If it is determined in step S03 that the supply of drive power based on the command profile is completed for the entire period of the command profile, the acquisition of the response profile is completed.
[0056] (Control parameter generation procedure) 4, the control device 100 first executes steps S11, S12, S13, S14, and S15. In step S11, the first model generation unit 114 generates a gain profile representing the relationship between the frequency of the command and the response gain of the control target 2 to the command, based on the command profile and the response profile, and stores the gain profile in the gain profile storage unit 115. In step S12, the first model generation unit 114 generates a first model based on a linear model in which the frequency characteristic is fitted to the gain profile.
[0057] In step S13, based on the command profile stored in the command profile storage unit 111 and the first model and the second model stored in the model storage unit 121, the model evaluation unit 122 generates a virtual response profile that simulates a time change in the state of the control target 2 according to the command profile, and stores the virtual response profile in the virtual profile storage unit 123. Note that before the second model generation unit 132 generates the initial second model, only the first model is stored in the model storage unit 121. In this case, the model evaluation unit 122 generates a virtual response profile based only on the command profile and the first model.
[0058] In step S14, the model evaluation unit 122 evaluates the error of the virtual response profile with respect to the response profile based on the response profile stored in the response profile storage unit 113 and the virtual response profile stored in the virtual profile storage unit 123. In step S15, the second model generation unit 132 checks whether the evaluation result of the error in step S14 is equal to or lower than the first level.
[0059] If it is determined in step S15 that the evaluation result of the error is not equal to or lower than the first level, the control device 100 executes steps S21, S22, S23, S24, and S25. In step S21, the second model generation unit 132 provisionally generates a second model based on a weighted sum of multiple types of unit models, and stores the second model in the model storage unit 121.
[0060] In step S22, the second model generation unit 132 generates an error model based on the provisionally generated second model and an error evaluation result based on the provisionally generated second model. For example, the second model generation unit 132 stores a record in which the provisionally generated second model and the error evaluation result based on the provisionally generated second model are associated with each other in the learning data storage unit 133. Thereafter, the second model generation unit 132 generates an initial error model based on the record stored in the learning data storage unit 133, and stores the generated initial error model in the error model storage unit 134.
[0061] In step S23, based on the command profile stored in the command profile memory unit 111 and the first model and the second model stored in the model memory unit 121, the model evaluation unit 122 generates a virtual response profile that simulates the change over time in the state of the control object 2 according to the command profile, and stores the virtual response profile in the virtual profile memory unit 123.
[0062] In step S24, the model evaluation unit 122 evaluates the error of the virtual response profile with respect to the response profile based on the response profile stored in the response profile storage unit 113 and the virtual response profile stored in the virtual profile storage unit 123. In step S25, the second model generation unit 132 checks whether the evaluation result of the error in step S24 is equal to or lower than the second level.
[0063] If it is determined in step S25 that the error evaluation result is not equal to or lower than the second level, the control device 100 executes steps S26 and S27. In step S26, the second model generation unit 132 changes the weighting based on the target error equal to or lower than the second level and the error model, and stores the second model after the weighting has been changed in the model storage unit 121.
[0064] In step S27, the second model generation unit 132 updates the error model based on the second model after the weight change and the error evaluation result based on the second model after the weight change. For example, the second model generation unit 132 accumulates a record in which the second model after the weight change and the error evaluation result based on the second model after the weight change are associated with each other in the learning data storage unit 133. Then, the second model generation unit 132 updates the error model in the error model storage unit 134 based on the record accumulated in the learning data storage unit 133. Then, the control device 100 returns the process to step S23. Thereafter, the change of the second model based on the error model and the update of the error model based on the changed second model are repeated until the error evaluation result becomes equal to or lower than the second level.
[0065] When it is determined in step S25 that the evaluation result of the error is equal to or lower than the second level, the control device 100 executes step S28. In step S28, the first model generation unit 114 checks whether the evaluation result of the error in step S24 is equal to or lower than the first level.
[0066] If it is determined in step S28 that the evaluation result of the error is not equal to or lower than the first level, the control device 100 executes step S29. In step S29, the first model generation unit 114 changes the gain profile based on the command profile, the response profile, and the second model. After that, the control device 100 returns the process to step S02. As a result, the first model is regenerated based on the changed gain profile.
[0067] If it is determined in step S15 that the evaluation result of the error is equal to or lower than the first level, and if it is determined in step S28 that the evaluation result of the error is equal to or lower than the first level, the control device 100 executes step S31. In step S31, the parameter generating unit 141 generates the above-mentioned control parameters based on the first model and the second model stored in the model storing unit 121, and stores them in the parameter storing unit 142. This completes the procedure for generating the control parameters.
[0068] [Control Procedure] Next, a control procedure based on the control parameters will be illustrated. As shown in Fig. 5, the control device 100 executes steps S41, S42, S43, S44, S45, S46, and S47 in order. In step S41, the control unit 112 acquires a target position. For example, the control unit 112 may acquire the target position based on a predetermined target motion pattern of the driven object 4, or may acquire the target position from a higher-level controller. In step S42, the control unit 112 calculates the feedforward compensation value based on the target position and the control parameters stored in the parameter storage unit 142.
[0069] In step S43, the control unit 112 acquires information on the current position and current speed of the driven object 4 based on the detection value of the sensor 6. In step S44, the control unit 112 calculates a target speed based on the deviation between the target position and the current position. In step S45, the control unit 112 calculates a driving force command based on the deviation between the target speed and the current speed.
[0070] In step S46, the control unit 112 adds the feedforward compensation value to the driving force command to calculate a corrected command. In step S47, the control unit 112 generates driving power based on the corrected command and supplies it to the motor 3. The control unit 112 repeats the above procedure at a predetermined control period.
[0071] [Application example] The above-mentioned configuration can be applied to a control object having a plurality of motors and a drive object including a plurality of rigid bodies. For example, the above-mentioned configuration can be applied to a multi-joint robot as shown in FIG.
[0072] The robot 10 shown in Fig. 6 is a so-called six-axis vertical articulated robot, and includes a base 11, a rotating unit 12, arms 13, 14, and 15, and a tool holder 16. The base 11 is fixed to a floor, a table, or an AGV (Automated Guided Vehicle) in a work area. The rotating unit 12 is provided on the base 11 so as to rotate around a vertical axis 31.
[0073] Arm 13 is connected to rotating part 12 so as to swing about axis 32 intersecting (e.g., perpendicular to) axis 31. The intersection here includes a case where there is a twisted relationship such as a so-called three-dimensional intersection. Arm 14 is connected to the tip of arm 13 so as to swing about axis 33 parallel to axis 32. Arm 15 is connected to the tip of arm 14 so as to swing about axis 34 intersecting axis 33 along arm 14, and to swing about axis 35 intersecting (e.g., perpendicular to) axis 34.
[0074] The tool holding unit 16 is connected to the tip of the arm 15 so as to rotate about an axis 36 along the center of the arm 15. The tool holding unit 16 holds various tools 17, such as a hand for gripping a workpiece, a welding torch, a paint gun, or a screw tightening tool.
[0075] Motor 21 rotates rotating part 12 around axis 31, motor 22 swings arm 13 around axis 32, motor 23 swings arm 14 around axis 33, motor 24 rotates the tip of arm 14 around axis 34, motor 25 swings arm 15 around axis 35, and motor 26 rotates tool holding part 16 around axis 36.
[0076] In the robot 10, the tool holding unit 16 and the tool 17 are the drive targets of the motor 26. The arm 15, the tool holding unit 16 and the tool 17 are the drive targets of the motor 25. The tip of the arm 14, the arm 15, the tool holding unit 16 and the tool 17 are the drive targets of the motor 24. The arms 14, 15, the tool holding unit 16 and the tool 17 are the drive targets of the motor 23. The arms 13, 14, 15, the tool holding unit 16 and the tool 17 are the drive targets of the motor 22. The rotating unit 12, the arms 13, 14, 15, the tool holding unit 16 and the tool 17 are the drive targets of the motor 21.
[0077] When a controlled object has multiple motors, such as the robot 10, the control device 100 may execute the above-described procedure for generating control parameters and the procedure for controlling the motors based on the control parameters for each motor.
[0078] [Effects of this embodiment] As described above, the control device 100 includes a first model generation unit 114 that generates a first model representing at least a part of the response characteristics of the control object 2 to a command based on a command profile that represents a time change of a command for the control object 2 and a response profile that represents a time change of the state of the control object 2 according to the command profile; a second model generation unit 132 that generates a second model that represents characteristics of the response characteristics that are not included in the first model based on the command profile, the response profile, and the first model; and a parameter generation unit 141 that generates control parameters that define the relationship between the command and the control amount for causing the control object 2 to follow the command based on the first model and the second model.
[0079] After a first model representing the characteristics of the control target 2 is generated, a second model representing the characteristics not included in the first model is further generated based on the first model, and a simulation is performed based on the first model and the second model. In this way, by using two types of models, the control target 2 can be modeled more appropriately. This is therefore effective in improving the accuracy of the simulation.
[0080] The second model generation unit 132 may generate the second model based on a plurality of types of unit models prepared in advance. In this case, it is possible to easily model characteristics (e.g., nonlinear characteristics) that are difficult to model using the first model.
[0081] Each of the multiple types of unit models may be a model that represents a nonlinear characteristic with respect to a command. In this case, it is possible to easily model a nonlinear characteristic that cannot be modeled by a linear model.
[0082] The control device 100 may further include a model evaluation unit 122 that generates a virtual response profile by simulating the time change of the state of the controlled object 2 according to the command profile based on the command profile, the first model, and the second model, and evaluates the error of the virtual response profile with respect to the response profile, and the second model generation unit 132 may generate the second model by provisionally generating a second model based on a weighted sum of multiple types of unit models, and changing the weights of the multiple types of unit models in the second model based on the evaluation result of the error based on the second model, repeatedly until a predetermined condition is satisfied. In this case, it is possible to more easily model characteristics that are difficult to model using the first model.
[0083] The second model generation unit 132 may generate an error model that represents the relationship between the weights of the multiple types of unit models in the second model and the error evaluation result based on the provisionally generated second model and the error evaluation result based on the provisionally generated second model, change the weights based on a predetermined target error and the error model, and update the error model based on the second model after the weights have been changed and the error evaluation result based on the second model after the weights have been changed, repeating this process until a predetermined condition is satisfied to generate the second model. In this case, it is possible to more quickly model characteristics that are difficult to model using the first model.
[0084] The first model generation unit 114 may generate a gain profile representing the relationship between the frequency of the command and the response gain of the control target 2 to the command based on the command profile and the response profile, and generate the first model based on a linear model in which the frequency characteristic is fitted to the gain profile. In this case, the linear model can be fitted by frequency analysis, making it possible to easily and appropriately generate a model.
[0085] The first model generation unit 114 may regenerate the first model based on the command profile, the response profile, and the second model when the error evaluation result after the second model is generated exceeds a predetermined level. In this case, after the second model is generated, the first model is regenerated taking the second model into consideration, which is expected to further improve the accuracy of the first model. Therefore, more appropriate modeling is possible.
[0086] When the error evaluation result after the second model is generated exceeds a predetermined level, the first model generation unit 114 may change the gain profile based on the command profile, the response profile, and the second model, and fit the frequency characteristic of the linear model to the changed gain profile to regenerate the first model. In this case, by changing the gain profile based on the second model, the first model can be regenerated as easily and appropriately as in the initial generation.
[0087] When the error evaluation result after the second model is generated exceeds a predetermined level, the first model generation unit 114 may generate a correction profile by removing the response of the second model from the response profile based on the command profile, the response profile, and the second model, and change the gain profile based on the command profile and the correction profile. In this case, the gain profile can be easily changed.
[0088] The second model generation unit 132 may regenerate the second model based on the command profile, the response profile, and the first model when the error evaluation result after the first model is regenerated exceeds a predetermined level. In this case, after the first model is regenerated, the second model is regenerated taking the first model into consideration, so that it is expected that the accuracy of the second model will be further improved. Therefore, more appropriate modeling is possible.
[0089] The control device 100 further includes a model evaluation unit 122 that generates a virtual response profile that simulates the time change in the state of the controlled object 2 according to the command profile based on the command profile, a first model, and a second model, and evaluates an error of the virtual response profile against the response profile, and the second model generation unit 132 generates the second model by tentatively generating a second model and modifying the second model based on an error evaluation result based on the second model, repeating this process until a predetermined condition is satisfied, and the first model generation unit 114 may regenerate the first model based on the command profile, the response profile, and the second model if the error evaluation result after the second model is generated exceeds a predetermined first level.
[0090] The second model generation unit 132 may generate the second model by repeatedly changing the second model based on the error evaluation result based on the second model until the error evaluation result becomes equal to or lower than a second level that is higher than the first level.
[0091] The second model generation unit 132 may regenerate the second model based on the command profile, the response profile, and the first model when the error evaluation result after the first model is regenerated exceeds a first level.
[0092] The controlled object 2 has a driven object 4 and a motor 3 that drives the driven object 4, the command is a command for a driving force to be generated by the motor 3, and the state of the controlled object 2 may be the speed of the driven object 4.
[0093] The controlled object 2 has a driven object 4 and a motor 3 that drives the driven object 4, the command is a command for a driving force to be generated by the motor 3, and the state of the controlled object 2 may be the position of the driven object 4.
[0094] The change in command over time in the command profile may include an oscillation of the command.
[0095] Although the embodiments have been described above, the present disclosure is not necessarily limited to the above-described embodiments, and various modifications are possible without departing from the gist of the present disclosure. [Explanation of symbols]
[0096] 2... controlled object, 3... motor, 4... driven object, 100... control device (production device of control system), 114... first model generation unit, 122... model evaluation unit, 132... second model generation unit, 141... parameter generation unit.
Claims
1. a first model generation unit that generates a linear first model that represents at least a part of a response characteristic of the controlled object with respect to a command, based on a command profile that represents a time change of a command for the controlled object and a response profile that represents a time change of a state of the controlled object according to the command profile; a second model generation unit that generates a second model representing a nonlinear characteristic of the response characteristic that is not included in the first model, based on the command profile, the response profile, and the first model; a model evaluation unit that generates a virtual response profile by simulating a time change in a state of the control target corresponding to the command profile, based on the command profile, the first model, and the second model, and evaluates an error of the virtual response profile with respect to the response profile; a parameter generating unit that generates a control parameter that defines a relationship between a control target and the command for causing the controlled object to follow the control target, based on the first model and the second model; The second model generation unit provisionally generating the second model based on a weighted sum of a plurality of types of unit models prepared in advance; A production device for a control system that generates the second model by repeatedly changing weights of the multiple types of unit models in the second model based on an evaluation result of the error based on the second model until a predetermined condition is satisfied.
2. 2. The production device of the control system according to claim 1, wherein each of the plurality of types of unit models is a model that represents a nonlinear characteristic with respect to a command.
3. The second model generation unit is generating an error model representing a relationship between weights of the plurality of types of unit models in the second model and the error evaluation result based on the provisionally generated second model and the error evaluation result based on the provisionally generated second model; 3. The control system production device according to claim 1, wherein the second model is generated by repeating the steps of changing the weights based on a predetermined target error and the error model, and updating the error model based on the second model after the weights have been changed and an evaluation result of the error based on the second model after the weights have been changed, until a predetermined condition is satisfied.
4. The control system production device according to any one of claims 1 to 3, wherein the first model generation unit generates a gain profile representing a relationship between a frequency of the command and a response gain of the controlled object to the command based on the command profile and the response profile, and generates the first model based on a linear model in which a frequency characteristic is fitted to the gain profile.
5. 5. The control system production device according to claim 4, wherein the first model generation unit regenerates the first model based on the command profile, the response profile, and the second model when the error evaluation result after the second model is generated exceeds a predetermined level.
6. 6. The control system production device of claim 5, wherein, when the error evaluation result after the second model is generated exceeds a predetermined level, the first model generation unit changes the gain profile based on the command profile, the response profile, and the second model, and fits a frequency characteristic of the linear model to the changed gain profile to regenerate the first model.
7. 7. The control system production device according to claim 6, wherein, when an evaluation result of the error after the second model is generated exceeds a predetermined level, the first model generation unit generates a correction profile by removing the response of the second model from the response profile based on the command profile, the response profile, and the second model, and changes the gain profile based on the command profile and the correction profile.
8. The production device for a control system according to any one of claims 5 to 7, wherein the second model generation unit regenerates the second model based on the command profile, the response profile, and the first model when an evaluation result of the error after the first model is regenerated exceeds a predetermined level.
9. The production device for a control system according to any one of claims 1 to 4, wherein the first model generation unit regenerates the first model based on the command profile, the response profile, and the second model when an evaluation result of the error after the second model is generated exceeds a predetermined level.
10. 10. The production device of a control system according to claim 9, wherein the second model generation unit generates the second model by repeatedly changing the second model based on a result of the error evaluation based on the second model until the result of the error evaluation becomes equal to or lower than a second level higher than the level.
11. 11. The production device of a control system according to claim 9 or 10, wherein the second model generation unit regenerates the second model based on the command profile, the response profile, and the first model when the evaluation result of the error after the first model is regenerated exceeds the level.
12. The controlled object includes a driven object and a motor that drives the driven object, 12. The production device of the control system according to claim 1, wherein the command is a command for a driving force to be generated by the motor, and the state of the controlled object is a speed of the driven object.
13. The controlled object includes a driven object and a motor that drives the driven object, 13. The production device of the control system according to claim 1, wherein the command is a command for a driving force to be generated by the motor, and the state of the controlled object is a position of the driven object.
14. The production apparatus of a control system according to any one of claims 1 to 13, wherein the time change of the command in the command profile includes an oscillation of the command.
15. a first model generation unit that generates a linear first model that represents at least a part of a response characteristic of the controlled object with respect to a command, based on a command profile that represents a time change of a command for the controlled object and a response profile that represents a time change of a state of the controlled object according to the command profile; a second model generation unit that generates a second model representing a nonlinear characteristic of the response characteristic that is not included in the first model, based on the command profile, the response profile, and the first model; a model evaluation unit that generates a virtual response profile by simulating a time change in a state of the control target corresponding to the command profile, based on the command profile, the first model, and the second model, and evaluates an error of the virtual response profile with respect to the response profile; a parameter generating unit that generates a control parameter that defines a relationship between a control target and the command for causing the controlled object to follow the control target, based on the first model and the second model; The second model generation unit is provisionally generating the second model; repeating the process of modifying the second model based on the evaluation result of the error based on the second model until a predetermined condition is satisfied to generate the second model; A production device for a control system, wherein the first model generation unit regenerates the first model based on the command profile, the response profile, and the second model when the error evaluation result after the second model is generated exceeds a predetermined level.
16. generating a linear first model representing at least a part of a response characteristic of the controlled object with respect to a command based on a command profile representing a time change of a command for the controlled object and a response profile representing a time change of a state of the controlled object according to the command profile; provisionally generating a second model representing a nonlinear characteristic not included in the first model among the response characteristics based on a weighted sum of a plurality of types of unit models prepared in advance; generating a virtual response profile by simulating a time change in a state of the control target according to the command profile, based on the command profile, the first model, and the second model, and evaluating an error of the virtual response profile with respect to the response profile; generating the second model by repeating the process of changing weights of the plurality of types of unit models in the second model based on an evaluation result of the error based on the second model until a predetermined condition is satisfied; generating, based on the first model and the second model, control parameters that define a relationship between a control target and the command for causing the controlled object to track the control target.
17. generating a linear first model representing at least a part of a response characteristic of the controlled object with respect to a command based on a command profile representing a time change of a command for the controlled object and a response profile representing a time change of a state of the controlled object according to the command profile; provisionally generating a second model representing a nonlinear characteristic not included in the first model among the response characteristics; generating a virtual response profile by simulating a time change in a state of the control target according to the command profile, based on the command profile, the first model, and the second model, and evaluating an error of the virtual response profile with respect to the response profile; generating the second model by repeating the process of modifying the second model based on an evaluation result of the error based on the second model until a predetermined condition is satisfied; regenerating the first model based on the command profile, the response profile, and the second model if the error evaluation result after the second model is generated exceeds a predetermined level; generating, based on the first model and the second model, control parameters that define a relationship between a control target and the command for causing the controlled object to track the control target.
18. generating a linear first model representing at least a part of a response characteristic of the controlled object with respect to a command based on a command profile representing a time change of a command for the controlled object and a response profile representing a time change of a state of the controlled object according to the command profile; provisionally generating a second model representing a nonlinear characteristic not included in the first model among the response characteristics based on a weighted sum of a plurality of types of unit models prepared in advance; generating a virtual response profile by simulating a time change in a state of the control target according to the command profile, based on the command profile, the first model, and the second model, and evaluating an error of the virtual response profile with respect to the response profile; generating the second model by repeating the process of changing weights of the plurality of types of unit models in the second model based on an evaluation result of the error based on the second model until a predetermined condition is satisfied; and generating, based on the first model and the second model, control parameters that define a relationship between a control target and the command for causing the controlled object to track the control target.
19. generating a linear first model representing at least a part of a response characteristic of the controlled object with respect to a command based on a command profile representing a time change of a command for the controlled object and a response profile representing a time change of a state of the controlled object according to the command profile; provisionally generating a second model representing a nonlinear characteristic not included in the first model among the response characteristics; generating a virtual response profile by simulating a time change in a state of the control target according to the command profile, based on the command profile, the first model, and the second model, and evaluating an error of the virtual response profile with respect to the response profile; generating the second model by repeating the process of modifying the second model based on an evaluation result of the error based on the second model until a predetermined condition is satisfied; regenerating the first model based on the command profile, the response profile, and the second model if the error evaluation result after the second model is generated exceeds a predetermined level; and generating, based on the first model and the second model, control parameters that define a relationship between a control target and the command for causing the controlled object to track the control target.
Citation Information
Patent Citations
Process operation planning system
JP2003295906A
Controller of multi-input multi-output system
JP2007109103A
Non-linear characteristic reproducing apparatus, non-linear resistance reproducing apparatus and non-linear characteristic reproducing program storage medium
JP2007149124A
Robot simulator, robot system, and simulation method
JP2018134703A
