JP5017649B2 - Motion learning system and motion learning method - Google Patents

Description

  The present invention relates to a motion learning system and a motion learning method that enable extraction, storage, and reproduction of motion of, for example, an expert.

  Motion control is a technical field that has developed rapidly over the last decade. Conventional motion control in a closed environment focused on suppressing disturbances, but recently, new motion control based on human interaction, such as the spread of robots into society and the development of nursing care systems, is desired. ing. In particular, in recent years, there has been a demand for the establishment of skill extraction techniques for advanced human work support such as minimally invasive surgical treatment by robots and skill preservation of skilled workers.

  Many studies have been conducted so far to extract human skills based on image information. In addition, tasks based on force control are necessary for the extraction and reproduction of tasks that repeatedly perform contact operations with the environment, such as those using a virtual environment or those using hybrid position and force control. Etc. have been proposed. However, at present, a satisfactory force control system is not necessarily instrumented, and it is not easy to realize. Although force information is originally bidirectional information governed by the “law of action / reaction”, the current force control system controls the current applied to the so-called actuator, resulting in the result. Unilateral force control that controls the axial torque, and lacks a basic method for handling bilateral information that provides force feedback in response to external force applied to the actuator. Because.

  As a specific example, for example, in Patent Document 1, when the master mechanism unit recognizes writing information and writing information related to a writing operation on the inner surface of the plate, which is a workpiece, the controller recognizes the robot arm of the slave mechanism and the slave side plate. The system that reproduces the writing action by controlling the movement is disclosed, but since this is trying to extract and reproduce the motion only by the position information, it repeats the contact / contact operation with the environment. It is difficult to reproduce.

  It is known that acceleration control for controlling the control stiffness to zero without losing robustness is indispensable for realizing an operation involving contact with the environment. By introducing such a control system based on acceleration control, it is possible to realize robust force feedback control using an actual human force input (which is a disturbance for the actuator) as a command value. Furthermore, by constructing a bilateral force feedback control system based on the master / slave system, from force information, which is bidirectional information governed by the “law of action / reaction”, from human action force and environment It is possible to extract the reaction force of each.

Many methods have been proposed to date for bilateral control to sense remote haptic information at hand. However, the conventional bilateral control method cannot be directly expanded to share control of force information among more systems, including force feedback type that controls force in the master system and position control in the slave system. It is hard to say that it has been established. Therefore, the inventors of the present application have so far proposed a new bilateral control system based on acceleration control, and have succeeded in sharply reproducing force information in the real world (Japanese Patent Application No. 2006-57632). Specifically, by performing acceleration control between the master system and the slave system, it is possible to easily realize integration of force control and position control. Artificial reproduction is possible. Further, since the control method based on this acceleration control can be extended to an infinite dimensional robot system, it can be generalized as so-called multilateral control.
JP 2006-62052 A

  However, the motion learning system that realizes the multilateral control has the following problems.

  In multi-lateral control, one master operator (instructor) teaches a plurality of subordinate operators (trainers) who are in remote locations by the master system and slave system. It becomes possible. However, since the operation of the sub operator returns to the main operator, and this restricts the operation of the main operator, there has been a concern that sufficient operation cannot be taught to the sub operator. .

  The present invention has been made in view of the above-described problems, and the purpose of the present invention is to provide a motion that can sufficiently teach the movement of the master operator to the slave operator without the movement of the slave operator returning to the master operator. The object is to provide a learning device and a motion learning method.

In order to solve the above-mentioned problem, the motion learning system of the present invention is a motion learning system that reproduces the operation of the main operator extracted by the main operation system using the sub operation system, and the sub operation system includes the main operation system. A virtual output means for outputting the operation of the operator as a first acceleration response and a first equivalent acceleration corresponding to the operation force of the slave operator are added , and the first equivalent acceleration and the first acceleration reference value are input, and movably master system that outputs a second acceleration response, the従操second equivalent acceleration is applied in accordance with the reaction force from the environment with the author of the operation, the second equivalent acceleration and the second acceleration reference value enter a, and moveable slave system that outputs a third acceleration response, as the second equivalent acceleration and the first equivalent acceleration is not output to the virtual output means, said first acceleration response , The second acceleration response, the third acceleration response, the first equivalent acceleration, and the second equivalent acceleration, the first acceleration reference value to the master system, and the second acceleration reference to the slave system. Multi-lateral control means for calculating each value and controlling the acceleration of the master system and the slave system.

In this case, the multilateral control means calculates the sum of the first equivalent acceleration and the second equivalent acceleration as a force servo so that the sum of the first equivalent acceleration and the second equivalent acceleration becomes zero. by multiplying the control parameter of, calculating a reference value of acceleration x ^{· ·} _ics ^ref sum mode, as deviation between the second acceleration response and the first acceleration response is zero at a position dimension, A difference obtained by subtracting the second acceleration response from the first acceleration response is multiplied by a control parameter of a position regulator to calculate a first acceleration reference value x ^·· _id1 ^ref in a difference mode, and the second acceleration response The difference obtained by subtracting the third acceleration response from the second acceleration response is multiplied by the control parameter of the position regulator so that the deviation between the third acceleration response and the third acceleration response becomes zero in the position dimension. Second acceleration of The reference value x ^·· _id2 ^ref is calculated, and the first force in the i-th slave operation system is such that an assist force proportional to the deviation between the first acceleration response and the second acceleration response is generated in the master system. The acceleration reference value x ^·· _im ^ref and the second acceleration reference value x ^·· _is ^ref are respectively calculated by ^Equation 21 described later .

Also, further preferably comprises an assist force storage means for storing the assist force.

  More preferably, the virtual output means includes a storage unit that stores the operation information of the main operator extracted by the main operation system, and an output unit that outputs the first acceleration response from the operation information of the storage unit, Consists of.

In order to solve the above-described problem, the motion learning method of the present invention is a motion learning method for reproducing the operation of the main operator extracted by the main operation system using the sub operation system, and the sub operation system includes the main operation system. A virtual output means for outputting the operation of the operator as a first acceleration response and a first equivalent acceleration corresponding to the operation force of the slave operator are added , and the first equivalent acceleration and the first acceleration reference value are input, and movably master system that outputs a second acceleration response, the従操second equivalent acceleration is applied in accordance with the reaction force from the environment with the author of the operation, the second equivalent acceleration and the second acceleration reference value enter a, and moveable slave system that outputs a third acceleration response comprises, as the said first equivalent acceleration second equivalent acceleration is not output to the virtual output means, said first acceleration response , The second acceleration response, the third acceleration response, the first equivalent acceleration, and the second equivalent acceleration, the first acceleration reference value to the master system, and the second acceleration reference to the slave system. Multi-lateral control is performed to calculate the respective values and to control the acceleration of the master system and the slave system.

In the multilateral control in this case, the added value of the first equivalent acceleration and the second equivalent acceleration is set to a force servo so that the sum of the first equivalent acceleration and the second equivalent acceleration becomes zero . by multiplying the control parameter, to calculate the acceleration reference value x ^{· ·} _ics ^ref sum mode, as deviation between the first acceleration response and the second acceleration response is zero at a position dimension, the A difference obtained by subtracting the second acceleration response from the first acceleration response is multiplied by a control parameter of the position regulator to calculate a first acceleration reference value x ^·· _id1 ^ref in a difference mode, and the second acceleration response and the The deviation obtained by subtracting the third acceleration response from the second acceleration response is multiplied by the control parameter of the position regulator so that the deviation between the third acceleration response and the third acceleration response becomes zero in the position dimension. Second acceleration reference The reference value x ^·· _id2 ^ref is calculated, and the first force in the i-th slave operation system is such that an assist force proportional to the deviation between the first acceleration response and the second acceleration response is generated in the master system. The acceleration reference value x ^·· _im ^ref and the second acceleration reference value x ^·· _is ^ref are respectively calculated by ^Equation 21 described later .

Also, it is preferable to store the assist force to assist force storage means.

  Further, it is preferable that the virtual output means stores the operation information of the main operator extracted by the main operation system and then outputs the first acceleration response from the operation information.

According to the system of claim 1 and the method of claim 5 , in the slave operation system, multilateral control for sharing force / tactile information is performed among the virtual output means, the master system, and the slave system. Therefore, it is possible to train, for example, a slave operator using the slave operation system while simultaneously reproducing the movement of the master operator accompanied with the environment operated by the master operation system using one or more slave operation systems. It becomes possible. In addition, since the slave operating system operated by the slave operator has a system configuration based on the master system and the slave system, the operation involving contact with the environment that could not be realized by the position control-based motion learning system or motion learning method. Can be transmitted.

Furthermore, the multilateral control in the slave operation system is performed between the master system and the slave system, not the main operation system itself, but the virtual output means for outputting the operation of the main operator as the first acceleration response, the first equivalent acceleration or the second equivalent acceleration, that is configured to not output to the virtual output means. Therefore, no matter what force the subordinate operator operates the master system of the subordinate operating system, the operation of the main operator is not dominated by the operation of the subordinate operator. It is possible for one master operator to perform training for one or more slave operators at the same time without being affected by force. Therefore, it is possible to sufficiently teach the operation of the primary operator to the secondary operator.

According to the system of claim 2 and the method of claim 6 , the master system and slave system of the slave operating system can artificially reproduce the action and the reaction. In addition, the operation of the main operator can be accurately transmitted from the virtual output means of the slave operation system to the master system and the slave system. Further, when a difference occurs in the operation between the primary operator and the secondary operator, an assist force proportional to the deviation acts on the master system, and the training effect of the secondary operator can be enhanced.

According to the system of claim 3 and the method of claim 7 , it is possible to quantitatively evaluate, for example, the skill level of the technique by storing the total assist force generated in the slave operation system. become.

According to the system of claim 4 and the method of claim 8 , since the operation information of a plurality of main operators can be stored in a database in the virtual output means, the slave operation system is used as a so-called training simulator. Will also be possible.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. First, an example of a control system for a robust acceleration control system using a disturbance observer applied by the motion learning system of the present invention will be described with reference to FIG. Such an acceleration control system is indispensable for realizing an operation involving contact with the environment. In the following description, except for figures and mathematical expressions, the first-order differential is represented as “ ^• ” and the second-order differential is represented as “ ^•• ” for convenience, and written after the corresponding symbol. In addition, “^” representing an estimated value is additionally written before the corresponding symbol.

In the figure, reference numeral 1 denotes a movable actuator that is provided as a target for estimating a disturbance force F ^dis and converts energy into power. This is achieved by inputting a current value i to a drive source such as a linear motor. It moves to the position x. A position sensor 2 such as a linear encoder is attached to the actuator 1 in order to convert the actual position x into an electrical position signal and output it.

On the other hand, the position information type disturbance observer 3 for controlling the actuator 1 inputs a current value i which is a command value to the actuator 1 and a value obtained by pseudo-differentiating the position x by the pseudo-differentiator 4, that is, a response speed value. ^ X ^· is input, and an estimated value of the disturbance force is output from each of these values. In actuality, the estimated value is composed of the pseudo-differentiator 4 and computer software. The pseudo-differentiator 4 converts the position signal into an estimated speed signal. By performing pseudo-differentiation on the position x, it is possible to obtain a conversion output with suppressed sensitivity to noise. The disturbance observer 3 incorporates an inverse model equivalent to the actuator 1, converts the current i into a first signal in force unit, and the response speed ^ θ _m ^· from the first signal and the pseudo-differentiator 4. The inverse model unit 5 that outputs a third signal that is compared with the second signal obtained by differentiating the signal, and the inverse model unit 5 is formed by setting a cutoff frequency during differentiation in the inverse model unit 5. The low-pass filter 6 extracts the third signal of the low frequency band component from the unit 5 and outputs it as the estimated disturbance force ^ F ^dis . Reference ^numeral 7 denotes force-current conversion means for reversely converting the estimated disturbance force ^ F ^dis into a compensation current value I ^cmp in the same unit as the target current reference value I ^ref , and the compensation current value I ^cmp and the current The current value (current command value) i to be input to the actuator 1 is calculated by adding the reference value I ^ref with the adder 8. Thereby, in the apparatus which consists of each structure of FIG. 1, the disturbance force F ^dis added to the actuator 1 from the outside can be removed. Since the actuator 1 outputs a position signal, the output of the actuator 1 (acceleration x ^·· is integrated twice to output the actual position x) through a two-stage integrator (1 / s) for convenience. Yes.

The control system shown in FIG. 1 constructs a robust acceleration control system by removing the influence of various disturbances acting on the motor (that is, disturbance force F ^dis ) with the disturbance observer 3. Here, a linear motor is adopted as the drive source of the actuator 1. The disturbance force F ^dis acting on the motor is expressed by the following equation.

Here, each symbol described in FIG. 1 and Equation 1 will be described. I ^ref is a current reference value, K _t is a thrust constant, x is a motor position, and a suffix after the symbol. _n means a nominal value. The first term on the right-hand side of the equation 1 is the inertia variation, the second term thrust ripple due to variations in the thrust constant, F _c is the Coulomb friction of the third term, D _m x ^· a fourth term viscous friction forces, F ^{ext in} the fifth term represents external force.

When the current reference value I ^ref and the motor speed x ^· can be detected by the pseudo-differentiator 4 having the cutoff frequency g _pd , the disturbance force F ^dis defined by Equation 1 is a disturbance observer 3 expressed by the following equation: Is estimated through the first-order low-pass filter 6 as follows.

However, in the above equation 2, g * _dis can be expressed as follows.

The estimated disturbance force F ^ ^dis is converted into a compensation current value I ^cmp by the force-current conversion means 7 as shown in the following equation and fed back to construct a control system having a control system that is robust against the disturbance. can do.

FIG. 2 is an equivalent representation of the control system shown in FIG. The robust control system based on the disturbance observer 3 in FIG. 1 is an acceleration control system as shown in FIG. From the sensitivity function of the mathematical formula shown in the element 11 of FIG. 2, it can be seen that the influence of the disturbance force F _dis is removed by setting the cutoff frequency g _dis of the disturbance observer 3 large here. Further, not only a robust acceleration control system is realized by the disturbance observer 3, but also only the external force F _ext applied to the motor can be estimated without a force sensor. In the actuator 1, a difference between the acceleration reference value x ^{·· ref} and the estimated acceleration obtained by converting the estimated disturbance force F ^ ^dis from the element 11 by the force-acceleration converting means 13 is output as the acceleration x ^···. It is a subtractor that equivalently expresses what is done. Thus, a robust acceleration control system for the motor constituting the actuator 1 can be realized.

Here, the haptic feedback control is indispensable for acquiring and reproducing the haptic information in the real world, and is a basic technology of the real world haptics. In the haptic feedback system, the input is an external force F _ext that is directly applied to the motor, and the purpose is to perform a copying operation by instantaneously feeding back this.

System for realizing the haptic feedback control is realized by the actuator 1 comprises a double disturbance observer loop, the block diagram is represented as shown in FIG.

In the figure, reference numeral 15 denotes a motor of the actuator 1, and an inner first disturbance observer 16 for realizing robust acceleration control includes the disturbance observer 3, the pseudo-differentiator 4 and the force-current converting means 7 shown in FIG. It consists of. Further, the outer second disturbance observer 17 estimates only the external force to the motor 15 by using the identified friction model or the like (estimated external force F ^ext ), and reverses this with the inverter 18 to feed back. It becomes possible. 3, 19 is an external force derives the acceleration reference value - by the acceleration converting unit, which is calculated by multiplication of said estimated external force F ^ ^ext and tactile feedback gain C _f. Reference ^numeral 20 denotes acceleration-current conversion means for deriving a current reference value I ^ref , which is calculated by multiplying the acceleration reference value x ^{·· ref} by (M _n / K _tn ).

The estimated external force F ^ext is estimated by the second disturbance observer 17 through the first-order low-pass filter as follows:

The system of FIG. 3 can be expressed in an equivalently simplified manner as shown in FIG. Here, the input to the acceleration dimension controller 23 is an acceleration reference value x ^{· · ref} and the external force of the equivalent acceleration x ^{· · ext,} from equivalent acceleration x ^{· · ext} of the external force to the acceleration response values x ^{· · ref} By setting the transfer function to 1, force feedback by the acceleration dimension controller 23 is realized. The equivalent acceleration x ^{·· ext} of the external force is obtained by dividing the estimated external force F ^ext by the nominal mass Mn as shown in the following equation, which is realized by the force-acceleration converting means 24 of FIG.

  In this way, by incorporating a loop configuration with the double disturbance observers 16 and 17 as shown in FIG. 4, a sensorless force feedback control with a robust and zero control rigidity can be constructed, and the compatibility with humans and the environment is improved. High contact operation can be realized.

  Next, an embodiment of a motion learning system incorporating the above-described preferred system configuration will be specifically described. In this embodiment, a system for transmitting an expert's actions and training by distance learning will be described in detail. An operation system for instructing an operation and an operation from the operation system are reproduced at a remote place. The present invention can also be applied to any motion learning system in which the operated system is connected by communication means.

  In transmitting the motion of the skilled person, it becomes a problem how to extract force information when the actuator 1 comes into contact with the environment and reproduce it at a remote place. Because haptic information is different from visual and auditory information, it is governed by the “law of action / reaction” in the real world, so in order to realize extraction and reproduction, bilateral haptics between the master system and the slave system are used. Feedback is required.

FIG. 5 is a schematic diagram showing the concept of artificial realization of the “law of action / reaction” by such bilateral force feedback. In the figure, an operation system 31 that is a main operation system installed on the operator side includes a master system 32 and a slave system 33. Thus, the operator (skilled) "of action and reaction law" in the real world O is the force acting on the operating body S of the operator O applied to the master system 3 2, slave system 3 3 undergoes environmental E Can be separated from the reaction force to the operating body S, and force information can be extracted and reproduced at a remote place while the skilled person operates.

FIG. 6 specifically shows the configuration of all control systems in the operation system 31 on the side of an expert, that is, an expert trainer (instructor). In the figure, a master system 32 and a slave system 33 that constitute the operation system 31 of the trained trainer respectively include acceleration control means 34 and 35 for the actuator 1 described in FIG. Here, the input to the master system 32 is the equivalent acceleration x ^·· _1m ^ext of the operating force of the skilled trainer and the acceleration reference value x ^·· _1m ^ref , and the output is the acceleration response value x ^·· _1m . The input to the slave system 33 is the equivalent acceleration x ^·· _1s ^ext of the reaction force from the environment E and the acceleration reference value x ^·· _1s ^ref , and the output is the acceleration response value x ^·· _1s .

In order to realize bilateral force feedback in the operation system 31, the sum of the operation force F _1m of the skilled trainer and the reaction force F _1s from the environment E is controlled to 0, and the position x _{1m of the} master system 32 It is necessary to configure the master system 32 and the slave system 33 to simultaneously realize the deviation from the position x _1s of the slave system 33 to 0. When the master system 32 and the slave system 33 are controlled as shown in FIG. 4 by the respective acceleration control means 34 and 35, the target of bilateral force feedback can be expressed as the following equations.

In order to calculate the added value x ^{· ·} _1c ^ext formula in Formula 7, the operation system 31, and the equivalent acceleration x ^{· ·} _{1 m} ^ext input to the master system 31, ^& equivalent acceleration x to be input to the slave system 32 ^- an adder 36 for adding the _1s ^ext is provided. Similarly, in order to calculate the deviation x ^·· _1d of the equation (8), the operation system 31 receives the acceleration response value x ^·· _1m output from the master system 31 and the acceleration output from the slave system 32. A subtracter 37 for subtracting the response value x ^·· _1s is provided.

Thus, the bilateral force feedback target equation shown in Equations 7 and 8 is expressed by the sum mode and the difference mode between the master system 32 and the slave system 33. Since the sum mode and the difference mode are expressed in an orthogonal virtual space, the above-mentioned goal can be satisfied by designing the control system independently in each space based on the acceleration control. Note that the equivalent acceleration x ^·· _1m ^ext input to the master system 32 can be calculated from the operating force F _1m of the skilled trainer using the force-acceleration converting means 38 provided in the operating system 31 by the following equation.

Similarly, the equivalent acceleration x ^·· _1s ^ext input to the slave system 33 is calculated from the reaction force F _1s from the environment E by the following equation using the force-acceleration conversion means 39 provided in the operation system 31. it can.

The added value x ^·· _1c ^ext obtained by the adder 36 is the sum of the equivalent acceleration x ^·· _1m ^ext of the operating force of the skilled trainer and the equivalent acceleration x ^·· _1s ^ext of the reaction force from the environment E Yes, this is an input to the operation system 31 that realizes bilateral force feedback control. By controlling the added value x ^·· _1c ^{ext to} be 0, the “law of action / reaction” can be artificially realized. The acceleration reference value x ^·· _1c ^ref in the sum mode is obtained by ^{converting the} added value x ^·· _1c ^ext and the force servo Cf by the force servo controller 40 that controls the added value x ^·· _1c ^{ext to} be 0. Multiply and generate as follows:

On the other hand, in Equation 8, x ^·· _1m represents the acceleration response value of the master system 32, x ^·· _1s represents the acceleration response value of the slave system 33, and the symbol “→” represents the position obtained by the second-order integration. It shows that the deviation becomes zero. The acceleration response deviation x ^·· _1d between the master system 32 and the slave system 33 is controlled to 0 by the position regulator controller 41 constituting the position regulator C _p (s) in the difference mode as shown in the following equation. .

Here, x ^·· _1d ^ref indicates an acceleration reference value in the difference mode. In order to simultaneously realize the force servo Cf in the sum mode and the position regulator C _p (s) in the difference mode, the actual operation system 31 has the acceleration reference values x ^{· •} _1c ^ref , x ^·· _1d ^ref are integrated in the acceleration dimension as shown in the following equation.

In the above equation, x ^·· _1m ^ref on the left side is an acceleration reference value input to the master system 32, and this value is obtained by ^converting each acceleration reference value x ^·· _1c ^ref and x ^·· _1d ^ref as shown in FIG. It is generated by an adder 42 for adding and an arithmetic unit 43 for multiplying the value obtained by the adder 42 by 1/2. Also, x ^{· ·} _1s ^ref is acceleration reference value to be input to the slave system 33, this value is as shown in FIG. 6, subtracts the acceleration reference value x ^{· ·} _1d ^ref from the acceleration reference values x ^{· ·} _1c ^ref The value is generated by a subtracter 44 and an arithmetic unit 45 that multiplies the value obtained by the adder 44 by ½.

As described above, since each of the actuators 1 of the master system 32 and the slave system 33 is subjected to the robust acceleration control by the acceleration control means 34 and 35, the acceleration reference value x ^·・ _1m ^ref and x ^{・ ・} _1s ^ref are realized, and bilateral force feedback is possible. Then, such force response and position response to the master system 32 and slave system 33 can be stored in a storage means (not shown), and the operation of the skilled trainer can be saved as a time-series database. .

  Next, a distance learning system based on multilateral control realized by the present embodiment will be described. The distance learning system proposed here has a function in which one skilled trainer simultaneously trains the technique of “Takumi” to a plurality of trainees (trainers) existing in a remote place. In order to realize such a function, the operation of the skilled trainer is transmitted from the operation system 31 shown in FIG. 6 to the operated system of each trainee by multilateral control through communication means (not shown).

  In general, multilateral control that shares haptic / tactile information among multiple systems artificially realizes the “law of action / reaction” among multiple multiple systems in remote locations. Therefore, it is possible to reproduce the feeling that a plurality of operators operate one system. However, when a trained trainer trains many trainees, it is clear that if multi-lateral control is applied as it is, the signal from the trainer's operation becomes relatively weak and it becomes difficult to transfer the motion. . Therefore, in this embodiment, a distance learning system is realized by proposing a multi-lateral control having a new configuration as shown in FIG.

In the system proposed in FIG. 7, the trainer system 51, which is the main operation system, has the same configuration as the operation system 31 described above, and will not be described in detail here. The sense of force of the skilled trainer is extracted by artificially reproducing the “law of action / reaction” by bilateral force feedback between the master system 32 and the slave system 33. On the other hand, each trainee system 52 ₁ to 52 _N corresponding to the operation reproducing system, based on the operation that has been extracted by the operation system 31 of the trainer, the virtual trainer 5 7, respectively. The N trainee systems 52 _{1 to} 52 _N are provided with a master system 55 and a slave system 56 similar to the trainer system 51, respectively. As a result, the “law of action / reaction” in the real world of the operator (trainer) O ′ is the force applied to the operating body S ′ of the operator O ′ applied to the master system 55 and the environment received by the slave system 56. It becomes possible to separate the reaction force from E ′ to the reaction force to the operating body S ′, and it becomes possible to extract the haptic information during the operation of the operator O ′.

In Figure 7, the master system 55 and the slave system 56 constituting the trainee system ₅₂ 1 to 52 _N, and between the virtual trainer 5 7 three systems, by constituting the multilateral control, Training system ₅₂ 1 - Regardless of the number of 52 _N , it is possible to reproduce the situation where the trained trainer trains each trainee on a one-on-one basis.

FIG. 8 shows the configuration of the entire control system in the i-th trainee system 52 _i shown in FIG. As described above, the trainee system 52 _i includes a virtual trainer 57 serving as a virtual output unit that outputs the acceleration response value x ^·· _1m from the master system 32 of the operation system 31, and a master system 55 including the acceleration control unit 61. The slave system 56 including the acceleration control system 62 is configured to mutually control in the acceleration dimension by the multilateral control means 63 described below. The actuators 1 of the master system 55 and the slave system 56 are subjected to robust acceleration control by the acceleration control means 61 and 62, respectively. These acceleration control means 61 and 62 are as described with reference to FIG. The input to the master system 55 is the equivalent acceleration x ^·· _im ^ext of the operating force of the ^trainee and the acceleration reference value x ^·· _im ^ref obtained by the multilateral control means 63, and the output is the acceleration response value x ^·· _{im im} It is. The input to the slave system 56 is the equivalent acceleration x ^·· _is ^ext of the reaction force from the environment E ′ and the acceleration reference value x ^·· _is ^ref , and the output is the acceleration response value x ^·· _is .

The equivalent acceleration x ^·· _im ^ext to the master system 55 can be calculated by using the force-acceleration conversion means 65 and dividing the nominal value M _n of the motor mass from the operating force F _im of the trainee. Similarly, the equivalent acceleration x ^·· _is ^ext to the slave system 56 _is obtained by dividing the nominal value M _n of the motor mass from the reaction force F _is from the environment E ′ using the force-acceleration conversion means 66. It can be calculated.

Here, assuming that the trainer system 51 is the first system and the trainee systems 52 _{2 to} 52 _N are the second and subsequent systems, each traini system 52 _{2 to} 52 _N is similar to the operation system 31 of the skilled trainer. A control system based on acceleration control is constituted by acceleration control means 61 and 62 provided in the master system 55 and the slave system 56, respectively. Based on this acceleration control, the target of the multilateral control to be introduced into each of the trainee systems 52 _{2 to} 52 _N is expressed by the following equations.

First, in Equation 14, the equivalent acceleration x ^·· _im ^ext of the operating force F _im of the trainee in each i (i = 2 to N) -th traini system 52 _{i and} the equivalent acceleration of the reaction force F _is from the environment E ′. By controlling the sum x ^·· _ic ^ext with x ^·· _is ^ext to 0, it is possible to artificially realize the “law of action / reaction”. This is the multilateral control unit 63, and the equivalent acceleration x ^{· ·} _im ^ext to the master system 55, an adder 68 for adding the equivalent acceleration x ^{· ·} _IS ^ext to the slave system 56, the adder 68 The sum mode control means 69 for outputting the acceleration reference value x ^·· _ic ^ref so that the added value x ^·· _ic ^ext obtained in step 1 becomes 0 is achieved. Here, the acceleration reference value x ^·· _ic ^ref in the sum mode is generated by the force servo controller 69 by multiplying the added value x ^·· _ic ^ext and the force servo Cf by the following equation. .

In the above equations 15 to 17, x ^·· _im is an acceleration response value of the master system 55 in each i (i = 2 to N) th trainee system 52 _i , and x ^·· _is Is an acceleration response value of the slave system 56 in the i (i = 2 to N) -th trainee system 52 _i . Here, the three multilateral system (i.e., the master system 55, the slave system 56, the virtual trainer 57) constituting the trainee system 52 _i during, the deviation ^{x · ·} each acceleration response _id1, ^{x · ·} _id2 , X ^·· _id3 is controlled to 0, so that the operation of the skilled trainer can be transmitted.

To enable operation transfer of such skilled trainer, multilateral control unit 63 shown in FIG. 8, the acceleration response values x ^{· ·} _im from the master system 55 from the acceleration response values x ^{· ·} _{1 m} from the virtual trainer 57 a subtracter 71 for calculating a deviation x ^{· ·} _id1 minus, the by the position regulator controller 72 which calculates an acceleration reference value x ^{· ·} _id1 ^ref as the deviation x ^{· ·} _id1 becomes zero at the position dimension 2 provided with a loop, a subtracter 73 for calculating a deviation x ^{· ·} _id2 by subtracting the acceleration response values x ^{· ·} _iS from the slave system 56 from the acceleration response values x ^{· ·} _im from the master system 55, ^- the deviation x ^· _id2 is provided with a third loop by the position regulator controller 74 which calculates an acceleration reference value x ^{· ·} _id2 ^ref such that 0 at position dimension. That is, each of the deviations x ^·· _id1 , x ^·· _id2 of the acceleration response in the multilateral system includes the position regulator controllers 72 and 74 having the position regulator C _p (s) in the difference mode as shown in the following equation. As a result, both are controlled to zero.

Here, x ^·· _id1 ^ref is an acceleration reference value in the difference mode from the position regulator controller 72, and x ^·· _id2 ^ref is an acceleration reference value in the difference mode from the position regulator controller 74. The difference mode between the three multilateral systems is determined by controlling one of the two difference modes, so that the remaining one difference mode is subordinately determined. Therefore, two loops including the position regulator controllers 72 and 74 are provided. Is sufficient.

As in the case of the bilateral force feedback, the multilateral control means 63 here also uses the respective acceleration reference values generated by the force servo Cf in the sum mode and the position regulator C _p (s) in the difference mode. x ^·· _id1 ^ref and x ^·· _id2 ^ref are integrated in the acceleration dimension as shown in the following expression, and the acceleration reference value x ^·· _im ^ref to the master system 55 which is the operation system of each trainee and the slave system 56 _Is configured to generate an acceleration reference value x ^·· _is ^ref .

In the above formula, the acceleration reference value x ^{· ·} _im ^ref to the master system 55, the acceleration reference values x ^{· ·} _{i c} ^ref from the force servo controller 69, the acceleration reference value x ^{· ·} from the position regulator control 72 It is generated by a subtractor 75 that subtracts _{i d1} ^ref and an arithmetic unit 76 that multiplies the value obtained by the adder 42 by ½. The acceleration reference value x ^·· _is ^ref to the slave system 56 _is ^derived from the acceleration reference value x ^·· _{i c} ^ref from the force servo controller 69 and the acceleration reference value x ^·· _{i d2} from the position regulator controller 74. It is generated by a subtractor 77 that subtracts ^ref, and an arithmetic unit 78 that multiplies the value obtained by the adder 42 by ½.

As described above, the trainee system 52 _i shown in FIG. 8 controls a plurality of trainees online at remote locations by controlling the multilateral systems including the virtual trainer 57 with the multilateral control means 63. Training can be performed simultaneously. Furthermore, the operation of the trainer trained in the database may be stored and used as the virtual trainer 57 in the trainee system 52 _i . By doing so, the trainee system 52 _i shown in FIG. 8 can be used as a training simulator without using an online communication means with the operation system 31.

More importantly, the multilateral control means 63 shown in FIG. 8 is arranged so that the acceleration response value x ^·· _1m is unilaterally output from the virtual trainer 57 without interference with the operation of the trainee, that is, The equivalent acceleration x ^·· _im ^ext of the operating force F _im and the equivalent acceleration x ^·· _is ^ext of the reaction force F _is from the environment E 'at that time are not output to the virtual trainer 57. That's what it means. In this way, no matter what force the trainee manipulates the master system 55 of the trainee system 52 _i , the operation of the trainer is not governed by the operation of the trainee. Without being affected, one skilled trainer can simultaneously train one or more trainees.

  Furthermore, conventional position control based training forces all the position responses in the trainer's operating system and trainee's operating system to match, forcing the movements to be synchronized independently of the trainee's operation. The training effect was not high. In the system proposed in the present embodiment, by introducing the multilateral control by the multilateral control means 63, when a difference occurs between the operation of the skilled trainer and the operation of the trainee, the assist force proportional to the deviation is generated. However, it works for the master system 55 in the second loop. Therefore, the training effect can be improved without forcibly synchronizing the operation regardless of the operation of the trainee.

FIG. 9 shows an example of an experimental apparatus for confirming the effectiveness of the distance learning system based on the multilateral control proposed in the above embodiment. As shown in this figure, the distance learning system for experiments is composed of one skilled trainer operating system 31 and a plurality of trainee operating systems (two in the figure), that is, trainee systems 52 ₁ and 52 _2. Is done. The operation system 31, together have a system configuration based on the master system 32 and the slave system 33 locally Training system 52 _1, 52 ₂ Similarly, based on the master system 55 and the slave system 56 at the local system It has a configuration.

Each of the actuators 1 of the experimental apparatus is composed of a linear motor 81 and corresponds to an operation performed with a human finger. The operating force of the operators O and O ′ to the linear motor 81 and the reaction force from the environments E and E ′ to the linear motor 81 are both sensed by the disturbance observers 16 and 17 as shown in FIG. Estimation can be performed without a sensor. The master system 32 and the slave system 33 constituting the operating system 31, together with includes a linear encoder 82 as a position detecting means of the actuator 1, respectively, Training system ₅₂ 1, 52 ₂ of the master system 55 and the slave system 56 In addition, a linear encoder 83 is provided as position detecting means for the actuator 1. Based on the detection signals from the linear encoders 82 and 83, the response values of the position, velocity, and acceleration described above can be obtained.

  The accuracy of friction identification used in the disturbance observers 16 and 17 for realizing sensorless force feedback strongly influences the force estimation accuracy. Therefore, in order to minimize the estimation error, the mechanical friction is reduced. This is very important. Here, as a rod-type linear motor 81 capable of mechanically removing the influence of friction, for example, a shaft motor manufactured by DMC Hillstone was used. The details of the linear actuator are shown in Table 1. Each part other than the actuator 1 is configured by a control program, and the sampling time is set to 100 μs.

In the experimental apparatus, an experiment was conducted in which an experienced trainer operating the master system 32 of the operation system 31 transmits the contact operation to the environment E to each of the trainee systems 52 ₁ and 52 ₂ . Initial position of the operation system 31 of the skilled trainer Training system 52 _1, 52 ₂ has an actuator 1 of each of the slave system 33,56, it is matched by installing in contact environment E, and E '. The operation for training is as follows.
(1) 1 second after the start of the experiment, 3 reciprocations of free motion (operation not in contact with the environments E and E ′) are performed for 3 seconds.
(2) Four seconds after the start of the experiment, the pushing operation to the environments E and E ′ is performed for 2 seconds.
(3) Six seconds after the start of the experiment, the operation moves from the contact operation to the environments E and E ′ to the free motion.

  Table 2 shows the control parameters used in the experiment.

10a to 10d show the experimental results. FIG. 10a shows the force response of the master system 32 and the slave system 33 in the operation system 31 of the trained trainer. FIG. 10b shows the same master system 32 and slave. Each position response of the system 33 is shown. Figure 10c is a Training system 52 ₁ in the master system 55, shows each force response of the slave system 56 and virtual trainer 57, and FIG. 10d is the same master system 55, indicates each position response of the slave system 56 and virtual trainer 57 ing.

  10a and 10b, since the operation system 31 is based on the master system 32 and the slave system 33, it is possible to separate the action force of the operator O and the reaction force from the environment E. It can be seen that the motion including the contact is accurately extracted. Furthermore, the sum of the action force of the operator O extracted by the master system 32 and the reaction force from the environment E extracted by the slave system 33 is almost zero, and the position between the master system 32 and the slave system 33 is Since the responses are also consistent, it can be confirmed that the “law of action / reaction” in the remote place can be reproduced well.

On the other hand, FIGS. 10c and Fig. 10d, it can be seen that can be realized bilateral force feedback also in the Training system 52 _1. In addition, as shown in FIG. 10d, the position response of the operator O ′ as a trainee and the position response of the virtual trainer 57 have a deviation around 4 seconds after the start of the experiment. This is a position regulator in the difference mode. It can be confirmed from FIG. 10 c that the assisting force to correct this is acting on the master system 55 by the controller 72. This assist force is not a feedback force when the slave system 56 comes into contact with the environment E ′. Therefore, when the vicinity of 4 seconds in FIG. 10c is seen, the force response of the slave system 56 becomes zero. Yes. Similarly, with respect to the operation deviation of the trainer (operator O) and the trainee (operator O ′) occurring from 6 seconds to 9 seconds, the position regulator controller It can be seen that the assist force is generated by 72 acting as a virtual spring / virtual damper, and the operation of the trainee is corrected.

  As described above, it was confirmed that each trainee can be optimally trained for the operation including the contact with the environments E and E ′ by the distance learning system proposed in this embodiment.

In the main operator or operating of the operator O and extracted with operating system 31, a motion learning system and motion learning how to reproduce with Training system 52 _i, Training system 52 _i this embodiment as described above, A virtual trainer 57 that outputs the motion of the operator O as a first acceleration response x ^·· _1m , and a first equivalent acceleration x ^·· _im ^ext corresponding to the manipulation force of the operator O 'as a slave operator are added , enter the first equivalent acceleration x ^{· ·} _im ^ext a first acceleration reference value x ^{· ·} _im ^ref, a moveable master system 55 you output a second acceleration response x ^{· ·} _im, the operator O ' a second equivalent acceleration x ^{· ·} _iS ^ext applied in accordance with the reaction force from the environment E 'due to the operation of the input of the second equivalent acceleration x ^{· ·} _iS ^ext and second acceleration reference values x ^{· ·} _iS ^ref to, the movable Allowed you output the third acceleration response _x ^·· is A slave system 56 comprises a, as further first equivalent acceleration x ^{· ·} _im ^ext and second equivalent acceleration x ^{· ·} _IS ^ext is not output to the virtual trainer 57, first acceleration response x ^{· ·} _{1 m,} The first acceleration reference value to the master system 55 from the second acceleration response x ^·· _im , the third acceleration response x ^·· _is , the first equivalent acceleration x ^·· _im ^ext , and the second equivalent acceleration x ^·· _is ^ext x ^·· _im ^ref and the second acceleration reference value x ^·· _is ^ref to the slave system 56 are calculated, respectively, and multilateral control means 63 for controlling the acceleration of the master system 55 and the slave system 56 is provided.

In this case, in place of the multilateral control means 63 , the first acceleration response x ^·· _{1m is set} so that the first equivalent acceleration x ^·· _im ^ext and the second equivalent acceleration x ^·· _is ^ext are not output to the virtual trainer 57. , Second acceleration response x ^·· _im , third acceleration response x ^·· _is , first equivalent acceleration x ^·· _im ^ext , and second equivalent acceleration x ^·· _is ^ext , refer to the first acceleration to the master system 55 By adopting a method of performing multilateral control for calculating acceleration of the master system 55 and the slave system 56 by calculating the value x ^·· _im ^ref and the second acceleration reference value x ^·· _is ^ref to the slave system 56 , respectively. Also good.

As a result, in the trainee system 52 _i , multilateral control for sharing force / tactile information among the virtual trainer 57, the master system 55, and the slave system 56 is performed. For this reason, the technique of “Takumi” is applied to, for example, the operator O ′ who uses the trainee system 52 _i while simultaneously reproducing the movement of the operator O operated by the operation system 31 by one or more trainee systems 52 _i. It is possible to carry out distance education training to pass on. Further, since the trainee system 52 _i operated by the operator O ′ has a system configuration based on the master system 55 and the slave system 56, contact with the environment E ′ that cannot be realized by the position control-based motion acquisition system is performed. The accompanying motion can be transmitted.

Further, the multilateral control in the trainee system 52 _i is not the main operation system itself, but the virtual trainer 57 that outputs the operation of the operator O as the first acceleration response x ^·· _1m , the master system 55 and the slave system 56. In addition, the equivalent acceleration x ^·· _im ^ext of the operating force F _im of the ^trainee and the equivalent acceleration x ^·· _is ^ext of the reaction force F _is from the environment E 'at that time are output to the virtual trainer 57 that is configured so as not to be. Therefore, even if the trainee operates the master system 55 of the trainee system 52 _i with any force, the operation of the trainer is not governed by the operation of the trainee, and thus is not affected by the force from the trainee. In addition, since one skilled trainer can simultaneously train one or more trainees, the operation of the operator O ′ does not return to the operator O, and the operation of the operator O can be performed. it is possible to sufficiently taught Training system 52 _i side of the operator O '.

In the multilateral control means 63 or multilateral control of the present embodiment, the sum of the first equivalent acceleration x ^·· _im ^ext and the second equivalent acceleration x ^·· _is ^ext is 0 in the i-th slave operation system. The sum mode acceleration is obtained by multiplying the added value x ^·· _ic ^ext of the first equivalent acceleration x ^·· _im ^ext and the second equivalent acceleration x ^·· _is ^ext with the control parameter Cf of the force servo so that The reference value x ^·· _ic ^ref is calculated, and the first acceleration so that the deviation x ^·· _id1 between the first acceleration response x ^·· _1m and the second acceleration response x ^·· _im becomes zero in the position dimension. response ^x from ^{· ·} _{1 m} deviation ^{x · ·} _id1 obtained by subtracting the second acceleration response ^{x · ·} _im, by multiplying the control parameter of the position regulator _C p and (s), the first acceleration reference value ^{x · ·} of the difference mode _id1 ^ref is calculated, and the deviation between the second acceleration response x ^·· _im and the third acceleration response x ^·· _{is Is} multiplied by the control parameter C _p (s) of the position regulator, so that the deviation x ^·· _id2 obtained by subtracting the third acceleration response x ^·· _is from the second acceleration response x ^·· _{im is set} to 0 in the position dimension. Te, calculates a second acceleration reference values x ^{· ·} _id2 ^ref difference mode, the assist force proportional to the deviation x ^{· ·} _id1 between the first acceleration response x ^{· ·} _{1 m} and the second acceleration response x ^{· ·} _im Are generated in the master system 55, the first acceleration reference value x ^·· _im ^ref and the second acceleration reference value x ^·· _is ^ref are respectively calculated by the above equation ( 21 ).

In this way, the master system 55 and the slave system 56 of the trainee system 52 _i can artificially reproduce the action and the reaction.

Also, the operation of the steering author O, it is possible to accurately transmit the virtual trainer 57 Training system 52 _i to the master system 55 and the slave system 56.

Further, when a difference occurs between the operations of the operators O and O ′, an assist force proportional to the deviation acts on the master system 55, and the training effect of the operator O ′ can be enhanced.

Further, although not shown in the figure, a configuration further including an assist force storage unit that stores the assist force or a method of storing the assist force in the assist force storage unit may be employed. By storing the sum of assist forces generated in the trainee system 52 _i , for example, it becomes possible to quantitatively evaluate the level of skill acquisition (skill level).

Although not shown, the virtual trainer 57 stores the operation information of the operator O extracted by the operation system 31, and the output for outputting the first acceleration response x ^·· _1m from the operation information stored in the storage unit. In the virtual trainer 57, the operation information of the operator O extracted by the operation system 31 is stored, and then the first acceleration response x ^·· _1m is output from the operation information. May be.

In this way, since the operation information of the plurality of operators O can be stored in the virtual trainer 57 as a database, the trainee system 52 _i can be used as a so-called training simulator.

  In addition, this invention is not limited to the said Example, A various deformation | transformation implementation is possible in the range of the summary of this invention.

  The motion learning system proposed in the present invention can be applied not only to a system operated by skilled trainers and trainees but also to any system in which remote operation is performed, but each has a system configuration based on a master / slave locally. As a result, it has become possible to transmit motion with contact with the environment that could not be realized with conventional position control-based systems. In addition, when there is a difference between the motion of the trainer trained as the primary operator and the motion of the trainee trained as the secondary operator, an assist force for correcting the motion is generated, so the training effect is dramatically improved. improves. Furthermore, the distance education based on such haptic information not only can improve the educational effect compared to the conventional image / sound based, but also integrates the operation of skilled trainers with the haptic information into a digital database. Can be an epoch-making solution to the problem of skill storage. It can be a basic technology for handling tactile media such as not only the development of production technology and advanced medicine based on the present invention but also the ability to transmit skills by broadcasting.

It is a block diagram showing an example of the robot system applicable to the motion acquisition system in one Example of this invention. FIG. 2 is a block diagram of an acceleration control system using a disturbance observer equivalent to FIG. It is a block diagram which shows an example of the system which implement | achieves force feedback control same as the above. FIG. 4 is a block diagram of a system equivalent to FIG. It is explanatory drawing which showed the concept of the artificial realization of the "law of action / reaction" by bilateral force feedback same as the above. FIG. 3 is a block diagram specifically showing the configuration of all control systems in the operation system on the expert trainer side. It is explanatory drawing which shows the conceptual structure of the distance education system based on multilateral control same as the above. FIG. 3 is a block diagram specifically showing the configuration of all control systems in the operation system of each trainee. It is explanatory drawing which shows the experimental apparatus for confirming the effectiveness of the distance education system based on multilateral control same as the above. It is an experimental result by the experimental apparatus of FIG. 9, Comprising: It is a graph which shows the force response in the operation system of an expert trainer. It is an experimental result by the experimental apparatus of FIG. 9, Comprising: It is a graph which shows the position response in the operation system of an expert trainer. FIG. 10 is a graph showing an experimental result of the experimental apparatus of FIG. 9 and a force response in the trainee system. FIG. FIG. 10 is a graph showing an experimental result by the experimental apparatus of FIG. 9 and showing a position response in the trainee system. FIG.

31 Operation system (main operation system)
51 Trainer system (main operation system)
52 _i Traini System (Subordinate Operation System)
55 Master system 56 Slave system 57 Virtual trainer (virtual output means)
63 Multilateral control means

Claims (8)

A motion acquisition system that reproduces the operation of the primary operator extracted by the primary operation system using the secondary operation system.
The slave operation system includes virtual output means for outputting the operation of the main operator as a first acceleration response;
A movable master system that receives a first equivalent acceleration according to the operating force of the slave operator, inputs the first equivalent acceleration and the first acceleration reference value, and outputs a second acceleration response;
A second equivalent acceleration corresponding to a reaction force from the environment accompanying the operation of the slave operator is added, and the second equivalent acceleration and the second acceleration reference value are input to output a third acceleration response. A slave system,
The first acceleration response, the second acceleration response, the third acceleration response, the first equivalent acceleration, and the second so that the first equivalent acceleration and the second equivalent acceleration are not output to the virtual output means. Multilateral control means for calculating the first acceleration reference value to the master system and the second acceleration reference value to the slave system from the equivalent acceleration, respectively, and controlling the acceleration of the master system and the slave system; , A motion acquisition system characterized by comprising. The multilateral control means is the i-th slave operation system,
A sum mode is obtained by multiplying the added value of the first equivalent acceleration and the second equivalent acceleration by the control parameter of the force servo so that the sum of the first equivalent acceleration and the second equivalent acceleration becomes zero. Acceleration reference value x ^·· _ic ^ref of
The deviation obtained by subtracting the second acceleration response from the first acceleration response is multiplied by the control parameter of the position regulator so that the deviation between the first acceleration response and the second acceleration response becomes zero in the position dimension. Then, the first acceleration reference value x ^·· _id1 ^ref in the difference mode is calculated,
The deviation obtained by subtracting the third acceleration response from the second acceleration response so that the deviation between the second acceleration response and the third acceleration response becomes zero in the position dimension is a control parameter of the position regulator. Multiply and calculate the second acceleration reference value x ^·· _id2 ^ref in the difference mode,
The first acceleration reference value x ^·· _im ^ref and the second acceleration reference value x ^·· _{is is} such that an assist force proportional to the deviation between the first acceleration response and the second acceleration response is generated in the master system. The motion learning system according to claim 1, wherein ^ref is calculated by the following formulas.
  The motion learning system according to claim 2, further comprising assist force storage means for storing the assist force.   The virtual output means includes a storage unit that stores the operation information of the main operator extracted by the main operation system, and an output unit that outputs the first acceleration response from the operation information of the storage unit. The motion acquisition system according to any one of claims 1 to 3, wherein A motion learning method for reproducing the operation of the master operator extracted by the master operation system using the slave operation system,
The slave operation system includes virtual output means for outputting the operation of the main operator as a first acceleration response;
A movable master system that receives a first equivalent acceleration according to the operating force of the slave operator, inputs the first equivalent acceleration and the first acceleration reference value, and outputs a second acceleration response;
A second equivalent acceleration corresponding to a reaction force from the environment accompanying the operation of the slave operator is added, and the second equivalent acceleration and the second acceleration reference value are input to output a third acceleration response. A slave system,
The first acceleration response, the second acceleration response, the third acceleration response, the first equivalent acceleration, and the second so that the first equivalent acceleration and the second equivalent acceleration are not output to the virtual output means. The first acceleration reference value to the master system and the second acceleration reference value to the slave system are calculated from equivalent acceleration, respectively, and multilateral control is performed to control the acceleration of the master system and the slave system. A motion learning method characterized by this. In the multilateral control, in the i-th slave operation system,
A sum mode is obtained by multiplying the added value of the first equivalent acceleration and the second equivalent acceleration by the control parameter of the force servo so that the sum of the first equivalent acceleration and the second equivalent acceleration becomes zero. Acceleration reference value x ^·· _ic ^ref of
The deviation obtained by subtracting the second acceleration response from the first acceleration response is multiplied by the control parameter of the position regulator so that the deviation between the first acceleration response and the second acceleration response becomes zero in the position dimension. Then, the first acceleration reference value x ^·· _id1 ^ref in the difference mode is calculated,
The deviation obtained by subtracting the third acceleration response from the second acceleration response so that the deviation between the second acceleration response and the third acceleration response becomes zero in the position dimension is a control parameter of the position regulator. Multiply and calculate the second acceleration reference value x ^·· _id2 ^ref in the difference mode,
The first acceleration reference value x ^·· _im ^ref and the second acceleration reference value x ^·· _{is is} such that an assist force proportional to the deviation between the first acceleration response and the second acceleration response is generated in the master system. ^6. The motion learning method according to claim 5, wherein each ^ref is calculated by the following mathematical formula.
The motion learning method according to claim 6, wherein the assist force is stored in an assist force storage unit. In the virtual output means stores the operation information of the main operator extracted by the main control system, then any one of claims 5-7, characterized in that outputs the first acceleration response from the operation information The motion learning method described in one.