JP2013054629A - Control apparatus and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a control apparatus in which admittance control can be mitigated from becoming unstable.SOLUTION: A robot control apparatus 1 comprises virtual object motion calculation means 7 and position control means 8 within power control means 3 which controls an arm 10 including an actuator 12 generating a thrust, an encoder 15 and a force sensor 11. The virtual object motion calculation means 7 calculates a target position pby solving an equation of motion including parameters of a virtual mass m, virtual viscosity cand target power fas virtual power to be applied to a virtual object, predetermined for the virtual object having a mass which is smaller than a mass of the arm and is not 0, using a contact force (f) detected by the force sensor 11 as input. The position control means 8 calculates a trust (g) to be applied to the arm using proxy based sliding mode control with the target position pand a position (p) of the arm detected by the encoder 15 as input and defines the trust as a command value to the actuator 12.

Description

  The present invention relates to the technology of a control apparatus and method, and more particularly to a control apparatus and method for performing position control-based force control.

2. Description of the Related Art Conventionally, there is known a position control device that quickly moves a hand position of a manipulator to a target position while limiting a force by an actuator that holds the manipulator (see Patent Document 1). The control law incorporated in the position control device disclosed in Patent Document 1 is called proxy-based sliding mode control (PSMC). The physical interpretation of PSMC will be described with reference to FIG. The controller of the PSMC does not directly apply a force to the control target 902 such as the hand of the manipulator when the set target position 901 is p _d , but the proxy ( proxy) A force is applied via a pointless virtual object called 903 having no mass.

The proxy 903 is coupled to the controlled object 902 by a virtual coupling element 904 and moves in the virtual space together with the controlled object 902. In the illustrated virtual connection element 904, K represents a proportional gain, and B represents a differential gain. In the controller (in the calculator), the proxy 903, which is a proxy point of the control target 902, is controlled by the PD control 902. Connected to. The proxy 903 is also connected to an SMC (sliding mode controller) 905 based on a sliding mode control law. The proxy 903 receives a force by PD control and a force by SMC (sliding mode control) 905. These forces are the same force and are always in equilibrium. This is because there is no physical contradiction because the proxy 903 does not have mass. In PSMC, position p _s proxy 903 is estimated by the position p of the detected control object 902, by applying force to the proxy 903, the position control to move the proxy 903 to the target position p _d is performed. For details of PSMC, see “R. Kikuuwe and H. Fujimoto,“ Proxy-based sliding mode control for accurate and safe position control ”, Proc. Of the 2006 IEEE Int. Conf. On Robotics and Automation, 2006, p.26- 31 ”.

  In addition to a position control device that controls the position of a control target that indicates the hand of a manipulator or the like, a force control device that controls the force of the control target is also known. Control (position input / force output) for outputting force command values using position information as an input is said to be a control device that performs impedance-type control (impedance control). Control (force input / position output) for outputting position command values with force information as an input is said to be a control device for performing admittance type control (admittance control).

  Conventionally, in an admittance control control device, a control device using P control, PD control, PID control or the like for position control inside force control is generally used. That is, in position control-based force control, P control, PD control, and PID control are used for internal position control. By using the position control base, there is an advantage that the influence of the friction of the actual actuator can be suppressed.

  As an example, when such a control device for admittance control is applied to a robot arm that works in cooperation with a person, for example, to reduce the impact on the person when the robot arm collides with a person. Control is required so that the robot arm immediately generates a thrust force in a direction away from the person. In order to improve the responsiveness when the robot arm collides with a person in this way, a method for controlling the robot arm, which is a control object in the virtual space, is assumed in the virtual space as an internal control of the control device. It is done. Specifically, in the force control of the robot arm, it is conceivable to improve the responsiveness by setting the virtual mass or virtual viscosity virtually given to the robot arm to a small value other than 0.

JP 2007-102748 A

  However, in a conventional admittance control control apparatus using P control, PD control, PID control, or the like for internal position control, there is a limit to control for setting a virtual mass or the like to be given to the robot arm to be small. The reason for this is that force control is based on an equation of motion, so within force control, the acceleration that generates the command value for the position controller increases as the virtual mass set for the robot arm decreases. This is because the gain becomes substantially high and unstable. In other words, if the mass is set lightly, the robot arm will become unstable as it tries to behave as if it is light, even though it is actually heavy. That is, there is a limit to the reduction of the virtual mass set for the robot arm. When the reduction of the virtual mass reaches the limit, the contact force when the robot arm comes into contact with a person does not become a constant value with time, but vibrates. That is, such a control method has a problem that admittance control becomes unstable.

  Therefore, an object of the present invention is to provide a control device and method that can solve the above-described problems and can mitigate instability of conventional admittance control.

  In order to solve the above-mentioned problems, the present inventors have made various studies on admittance control. As a result, even if the reduction of the virtual mass set by the conventional control using P control, PD control, PID control, or the like is used as the position control that is the base of force control, PSMC ( We found that the instability of admittance control can be alleviated by using proxy-based sliding mode control.

  Therefore, the control device according to the present invention includes force control means having a position control means inside, and the force control means performs admittance control based on force information input from the outside, whereby an external object and A control apparatus for controlling a contact force generated with another object, wherein the position control means uses proxy-based sliding mode control.

  In the control device according to the present invention, the position control unit includes a feedback controller including a proportional control for performing feedback control including at least a proportional operation, and a sliding mode controller based on a sliding mode control law, A feedback controller including a control and the sliding mode controller cooperate to perform the proxy-based sliding mode control, and the feedback controller including the proportional control includes a point-like object having no mass in a virtual space. A virtual viscoelastic force applied to the proxy is calculated from a virtual connection element that virtually connects the proxy, which is a virtual object, and the external object, and the sliding mode controller includes the sliding mode controller, It is a force that balances the viscoelastic force, and the position of the proxy is It is preferable to calculate the virtual force applied to the proxy when performing control to converge to the target position set in the Gumodo controller as thrust applied to the object.

  Further, in an aspect in which the object includes a drive mechanism that generates a thrust applied to the object and a position detection sensor, the control device according to the present invention is configured so that the force control unit corresponds to the object. Parameters of a virtual mass, a virtual viscoelasticity, and a target force as a virtual force applied to the virtual object, which are predetermined for a virtual object having a predetermined non-zero mass smaller than a substantial amount of Virtual object motion calculation means for calculating a target position of the virtual object by solving a motion equation for the virtual object using the contact force as an input, and the position control means includes the calculated target position. And a position of the object detected by the position detection sensor as an input to a drive mechanism that drives the object using the proxy-based sliding mode control. It is preferable to output a click (thrust) command value.

  According to this configuration, the control device calculates the target position in the force control based on the equation of motion of the virtual object having a non-zero mass smaller than the substantial amount corresponding to the object, and applies the virtual force to the object. It can be calculated as a thrust applied to and output as a command value.

  In order to solve the above problems, a control method according to the present invention includes force control means having position control means therein, and the force control means performs admittance control based on force information input from the outside. A control method by a control device that controls a contact force generated between an external object and another object by performing the position control means using proxy-based sliding mode control.

  According to the present invention, destabilization of conventional admittance control can be mitigated.

It is a conceptual diagram which shows typically the human robot cooperation system containing the robot control apparatus which concerns on embodiment of this invention. It is a block diagram which shows typically the structure of the robot control apparatus which concerns on embodiment of this invention. It is a flowchart which shows the outline of the flow of force control by the robot control apparatus which concerns on embodiment of this invention. It is a block diagram which shows typically the structure of the verification apparatus which verifies the performance of the robot control apparatus which concerns on embodiment of this invention. It is a conceptual diagram which shows typically the control model in the virtual object motion calculator of a verification apparatus. It is a control block diagram of an output restriction type PD controller used as a comparative example for the position controller of the verification device. It is a control block diagram of PSMC used as an embodiment for the position controller of the verification device. It is a graph which shows an example of the verification result by a verification apparatus. It is a schematic diagram which shows the physical interpretation of the conventional PSMC controller.

  DESCRIPTION OF EMBODIMENTS Hereinafter, modes for carrying out the present invention (referred to as embodiments) will be described in detail with reference to the drawings.

[1. Configuration of human-robot collaboration system]
A human-robot collaboration system 100 shown in FIG. 1 includes a robot R and a robot control device 1. The robot control apparatus 1 is connected to the robot R via a cable 20 and controls the arm 10 so that the arm 10 of the robot R can work in cooperation with the person P. In the present embodiment, an object outside the robot control apparatus 1 will be described as the arm 10, but the present invention is not limited to this. In the example illustrated in FIG. 1, the robot R includes two arms 10, but the number of arms is not limited to this. As the arm 10, the arm 10 can be used in accordance with the work application. The arm 10 is constituted by, for example, an industrial robot arm having a multi-joint and multi-spindle axes.

As shown in FIG. 2, the arm 10 of the robot R mainly includes a force sensor 11 and an actuator 12. In FIG. 2, one arm 10 is illustrated.
One or more force sensors 11 are provided at the tip of the arm 10, for example. The force sensor 11 includes a sensor for detecting a pressing force necessary for work and a sensor for detecting contact with a person P or an obstacle. In FIG. 2, one force sensor 11 is illustrated.

  The actuator 12 is a drive mechanism that generates a thrust of the arm 10 and forms a joint or the like of the arm 10. The actuator 12 includes, for example, a servo motor 14 that generates a driving force for bending and expanding / contracting the joint and an encoder 15 as a position detection sensor. In FIG. 2, one actuator 12 is illustrated.

[2. Configuration of control device]
The robot control apparatus 1 includes, for example, a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), an HDD (Hard Disk Drive), an input / output interface, and the like, as shown in FIG. The target force setting means 2, the force control means 3, and the servo amplifier 4 are mainly provided.

The target force setting means 2 sets a target force that is a target of thrust of the arm 10 in the force control means 3. The set value of the target force may be input from the outside of the robot control device 1, for example, or may be stored inside the robot control device 1. For example, when it is stored internally, for example, a fixed target force corresponding to the work performed by the arm 10 can be stored, or a target force that changes with time can be programmed and stored.
For example, the target force f _d when the arm 10 collides with a person has a size that does not cause damage when the arm 10 comes into contact with another object and can ensure responsiveness, for example,
(Acceleration of virtual object) = (target force f _d −virtual viscosity × virtual object speed) / (mass of virtual object) ≧ 4.9 m / s ² .

The force control means 3 has a position control means 8 inside, and performs contact between the external object (arm) and another object by performing admittance control based on force information input from the outside. The force is controlled, and the calculated torque (thrust) command value is output to the servo amplifier 4.
The servo amplifier 4 controls the servo motor 14 by converting a torque (thrust) command value into a current command value and outputting the current to the servo motor 14.

In the present embodiment, the force control unit 3 includes a target force acquisition unit 5, a contact force acquisition unit 6, a virtual object motion calculation unit 7, and a position control unit 8, as illustrated.
The target force acquisition unit 5 acquires the target force from the target force setting unit 2 and is configured by a predetermined interface.
The contact force acquisition means 6 acquires the contact force detected by the force sensor 11 of the arm 10 and is composed of a predetermined interface.

  The virtual object motion calculation means 7 uses the contact force detected by the force sensor 11 as an input, and calculates the target position of the virtual object by solving the motion equation for the virtual object including a predetermined parameter. It is. The calculated target position is output to the position control means 8 as a target position command value. Here, the virtual object corresponds to the arm 10 in the virtual space corresponding to the arm (calculator) when it is assumed that the arm 10 has a predetermined mass that is smaller than the substantial amount of the arm and is not 0 and performs a predetermined acceleration motion. (Inside). Hereinafter, it is assumed that the virtual object does not mean a proxy. Note that the virtual object is introduced only in the force control by the virtual object motion calculation means 7 and is not introduced in the processing of the position control means 8, but it is assumed that it is apparently at the same position as the arm 10. Also good. The contact force can also be obtained by estimation (simulation) without using the detection value of the force sensor 11.

Virtual object motion calculator 7 is, by inputting the contact force f in equation (1) below that shows the dynamic characteristics of the virtual object, the equation (1) the left side of the target acceleration command value (position command value p _d Time 2 Second order differential value) is calculated at a predetermined operation cycle. This virtual object motion calculator 410 performs integration twice with respect to the time second-order differential value on the left side of Equation (1) to obtain the position command value p _d as the target position, and outputs it to the position control means 8. The position command value p _d is also expressed as a target position p _d .

Here, the parameter f _d is a virtual force (target force) applied to the virtual object, f is a force (contact force) applied to the contacted object such as a human, v is the speed of the virtual object, and m _d is the virtual object mass, c _d denotes the viscosity of the virtual object.
For example, f _d, m _d, keep impart a predetermined value as c _d.
The detection value of the force sensor 11 can be used as f.
v can be obtained, for example, by performing a calculation integrating the equation (1). Also, an actual measurement value of the speed of the arm 10 can be used.

In this embodiment, the target force f _d is input from the target force acquisition unit 5 and the contact force f is input from the contact force acquisition unit 6 to the virtual object motion calculation unit 7. The calculated position command value p _d is output to the position control means 8.

Here, the position in the position command value p _d is a target position of the virtual object, and indicates a target value of a position that can be detected by the encoder 15 of the actuator 12 (this is also simply referred to as the position of the actuator 12). ing. The position that can be detected by the encoder 15 may be the output value of the encoder, the rotational position of the servo motor 14 converted from the encoder output, and the position coordinates of the position where the arm 10 has moved. Hereinafter, the position detected by the encoder 15 is taken as the position of the arm 10 (referred to as a position p) as an example. Therefore, the position of the actuator 12 is used in the same meaning as the position of the arm 10 as an example. In this sense, the position of the arm 10 can be interpreted as the position of the virtual object.

  The position control means 8 outputs a torque (thrust) command value for driving the actuator 12 provided in the arm 10 using known proxy-based sliding mode control (hereinafter also referred to as PSMC).

  The position control means 8 includes, for example, a PD controller as the virtual connection element 904 shown in FIG. 9 and an SMC (sliding mode controller) 905 shown in FIG. 9, and the PD controller and SMC cooperate with each other. Perform proxy-based sliding mode control. In this case, the PD controller includes a virtual viscosity applied to the proxy from a virtual connection element that virtually connects the proxy, which is a point-like virtual object having no mass in the virtual space, and the arm 10. Elastic force is calculated by proportional action and differential action. The SMC is a force that balances the virtual viscoelastic force, and is a thrust that applies to the arm 10 a virtual force that is applied to the proxy when performing control to converge the position of the proxy to the target position set in the SMC. Calculate as

In the present embodiment, the position control means 8, together with the position command value p _d from the virtual object motion calculator 7 is inputted as a target position, the position p of the arm is detected by the encoder 15 of the actuator 12 is inputted. The position control means 8 calculates a torque (thrust) command for driving the actuator 12 using PSMC based on these inputs.

[3. Flow of force control by robot controller]
The flow of force control by the robot controller 1 will be described with reference to FIG. 3 (see FIG. 2 as appropriate). FIG. 3 is a flowchart showing an outline of a flow of force control by the robot control apparatus according to the embodiment of the present invention. Here, as an example, it is assumed that the target force f _d is set to a fixed value.

The robot control device 1 detects the contact force f by acquiring the output of the force sensor 11 by the contact force acquisition means 6 (step S1). When the contact force f is smaller than the target force f _d −c _d v (step S2: Yes), the robot control apparatus 1 uses the virtual object motion calculation unit 7 to accelerate the target position p _d to be deeper than the current position. Is set (step S3), and the target position _pd is set (updated) (step S7). The contact force of the arm 10 can be increased by setting the target position _pd to a position deeper than the current position.

In the determination of the contact force f, when the contact force f is larger than the target force f _d −c _d v (step S4: Yes), the robot control device 1 uses the virtual object motion calculation unit 7 to execute the target position p _d. Is set so as to be shallower than the present (step S5), and the target position _pd is updated (step S7). It is possible to reduce the contact force of the arm 10 by setting the target position p _d at a position shallower than the current.

In the determination of the contact force f, when the contact force f is equal to the target force f _d −c _d v (step S4: No), the robot controller 1 sets the acceleration to zero (step S6), and the target force The position _pd is updated (step S7).

Then, following the update of the target position p _d (step S7), and the robot controller 1, the position control means 8, calculates the torque (thrust) command value (step S8). Then, the robot controller 1 converts the torque (thrust) command value into a current command value by the servo amplifier 4, supplies the current to the actuator 12 of the arm 10 of the robot R (step S9), and returns to step S1. . Therefore, even if a large collision force is generated at the moment of the collision, the robot controller 1 repeats these processes, so that the contact force f and the target force f _d −c _d v eventually become equal. That is, it is possible to control so as to output a thrust in a direction away from the person P when the arm 10 is stopped.

[4. Robot controller performance]
Here, the performance of the robot control apparatus 1 when the arm 10 collides with a person will be described. The performance was verified by performing a collision test assuming that the arm 10 collides with a person. In the following, 4-1. Configuration of verification apparatus, 4-2. Control model, 4-3. Configuration of comparative example of position controller, 4-4. Configuration of embodiment of position controller, 4-5. Crash test conditions, 4-6. The test results will be described in detail.

<4-1. Configuration of verification device>
The configuration of the verification device used in the collision test will be described with reference to FIG. FIG. 4 is a block diagram schematically showing the configuration of a verification apparatus that verifies the performance of the robot control apparatus according to the embodiment of the present invention. As shown in FIG. 4, the verification device 400 includes an admittance controller 401, a servo amplifier 402, a linear motion actuator 403, a collision body 404, a force sensor 405, and a collision target 406 including a buffer material 407. .

  The admittance controller 401 outputs a torque (thrust) command for controlling the collision force by the collision body 404 by moving the position of the collision body 404 with the linear motion actuator 403 as a control target, and the virtual object motion A calculator 410 and a position controller 420 are provided.

The virtual object motion calculator 410 has the functions of the target force acquisition means 5, the contact force acquisition means 6 and the virtual object motion calculation means 7 shown in FIG.
The virtual object motion calculator 410 receives the target force f _d set for the virtual object in force control, and the contact force f between the collision object 404 and the collision object 406 detected by the force sensor 405. Enter. Here, the virtual object is an object on the virtual space (in the computer) corresponding to the collision object when it is assumed that the collision object 404 has a predetermined mass that is not 0 smaller than the actual mass. The virtual object motion calculator 410 calculates a position command value pd that is a target position of the linear actuator 403 according to a control model described later, and outputs the position command value p _d to the position controller 420.

The position controller 420 calculates a torque (thrust) command for driving the linear actuator 403 as a moving mechanism based on the position command value p _d input from the virtual object motion calculator 410. The position controller 420 receives the position command value p _d set as the target position of the linear actuator 403 and the position p detected by an encoder (not shown) of the linear actuator 403.

In order to compare the performance, a verification device in the case where the output limiting PD controller 601 shown in FIG. Details of the output limiting PD controller 601 will be described later.
In addition, a verification apparatus when the PSMC 701 shown in FIG. 7 is adopted as the position controller 420 is taken as an example. The PSMC 701 has the function of the position control means 8 shown in FIG. Details will be described later.

The servo amplifier 402 is the same as the servo amplifier 4 shown in FIG.
The linear actuator 403 has the same function as the actuator 12 shown in FIG. 2 and is a control target of the admittance controller 401. The linear motion actuator 403 is a moving mechanism that moves the collision body 404 and includes a servo motor, a ball screw, a linear motion guide, and an encoder.

The collision body 404 corresponds to the hand of the arm 10 shown in FIG. 2, includes a force sensor 405, and slides on a straight line when the linear actuator 403 is driven to collide with the collision target 406. FIG. 4 shows the state before the collision.
The force sensor 405 has the same function as that of the force sensor 11 shown in FIG. 2, detects the contact force when it comes into contact with the collision object 406, and feeds it back to the admittance controller 401.
A shock-absorbing material 407 made of a sponge is provided on the surface of the collision target 406 that contacts the collision body 404. The buffer material may be provided on the collision body, or may be provided on both.

<4-2. Control model>
A control model for calculating the target position in the virtual object motion calculator 410 will be described with reference to FIG. FIG. 5 is a conceptual diagram schematically showing a control model in the virtual object motion calculator of the verification device. The configuration shown in FIG. 5 is an example of admittance control.
The virtual object 501 is an object in the computer corresponding to the collision object when it is assumed that the collision object 404 shown in FIG. 4 has a non-zero mass m _d smaller than the actual mass in force control. FIG. 5 shows the state when the vehicle collides.

  The contacted object 502 is an object on the virtual space (in the computer) corresponding to the impacted object 406 including the cushioning material 407 shown in FIG. A force (contact force) applied to the contacted object 502 by the virtual object 501 is defined as f. Here, the force f is the same as the contact force of the collision body 404 when the collision body 404 shown in FIG. Therefore, the contact force f was measured using a force sensor 405 shown in FIG. Note that the contact force f may be obtained by estimation (simulation) without performing direct measurement by the force sensor.

The force generation element 503 corresponds to the target force f _d shown in FIG. 4 and represents a virtual force applied to the virtual object 501 (a force set for the virtual object 501).
Virtual damper 504 represents a damper in the virtual space (calculated flight) for generating a viscous c _d of the virtual object 501. In calculating the viscous resistance (c _d v), a value obtained by integrating the equation (1) was used as the velocity v of the virtual object 501.
A target position p _d of the virtual object 501 indicates a position command value p _d that is a target position of the collision body 404.

  Note that the actual speed of the collision object 404 may be used as the speed v of the virtual object 501. Moreover, it is not necessary to limit to the collision (contact) from a certain speed. Moreover, you may be in a contact state from the beginning.

The virtual object motion calculator 410 inputs the contact force f into the above-described equation (1) indicating the dynamic characteristics of the virtual object 501, thereby obtaining the target acceleration (the second order time of the position command value p _{d on} the left side of the equation (1)). a differential value) calculated at predetermined operating cycle, further integrating the calculated two runs, calculates a target position p _d of the virtual object, and outputs to the position controller 420.

<4-3. Configuration of Comparative Example of Position Controller>
A comparative example of a position controller that calculates a force command value will be described with reference to FIG. 6 (see FIG. 4 as appropriate). FIG. 6 is a control block diagram of an output limiting PD controller used as a comparative example for the position controller of the verification apparatus.
Output restriction type PD controller 601 calculates the force g applied to the controlled object 602 (impact body 404 in FIG. 4) on the basis of the position p of the controlled object 602 and the position command value p _d, a force command of it to the actuator Value. Here, the position p of the control object 602 is the position of the collision body 404, that is, the detected value of the encoder. In FIG. 6, h is an external force applied to this system, and SAT 603 represents a unit saturation function sat (x) shown in the following equation (2).

  The output-restricted PD controller 601 calculates the right side of Expression (3) as the control represented by the control block in the frame indicated by the broken line in FIG. 6, thereby controlling the control target 602 (the collision object 404 in FIG. 4). The applied force g is calculated.

Here, F represents the upper limit value of the force generated by the linear actuator 403 that is a moving mechanism. K is a proportional gain, B is a differential gain, p _d is a position command value, p is a position of the collision body 404, that is, an encoder detection value of the linear actuator 403, and sat (x) is a unit saturation function of the equation (2). Show.

<4-4. Configuration of Embodiment of Position Controller>
An embodiment of the position controller will be described with reference to FIG. 7 (refer to FIGS. 4 and 9 as appropriate). FIG. 7 is a control block diagram of the PSMC used as an embodiment for the position controller of the verification apparatus. As an example, the PSMC control block is realized by a combination of a PD controller (virtual connection element 904 in FIG. 9) and a sliding mode controller (SMC 905 in FIG. 9). The PSMC control block can also be realized by a combination of a P controller or PID controller and a sliding mode controller.

The PSMC 701 shown in FIG. 7 has a force command value to the actuator based on the position command value p _d (the target position 901 set in the SMC 905 in FIG. 9) and the position p of the control object 702 (the control object 902 in FIG. 9). g is calculated. Here, the position p of the control object 702 is a detection value of the encoder. Further, it is assumed that a proxy (proxy 903 in FIG. 9) is connected to the control object 702 (control object 902 in FIG. 9). A proxy is a point-like virtual object with no mass. The position of the proxy is p _s (see FIG. 9).

  In FIG. 7, h is an external force applied to this system, and SGN 703 represents a sign function sgn (x) shown in the following equation (4).

  According to the definition of Expression (4), sgn (x) = + 1 when x> 0, and sgn (x) = − 1 when x <0. Equation (4) defines that sgn (x) can take any value in the closed interval [-1, 1] when x = 0. That is, the SGN 703 illustrated in FIG. 7 receives the position information and outputs a dimensionless value from −1 to 1. Note that there is a relationship between the equation (4) and the equation (2) that y = sgn (xy) and y = sat (x) are equivalent when a predetermined condition is satisfied. It is known (see, for example, JP-A-2008-97152).

  The PSMC 701 shown in FIG. 7 is a control expressed mainly by the lower control block in the frame indicated by a broken line in FIG. As a control of the block, the lower equation of equation (5) is solved to calculate the force g to be applied to the controlled object, and this is used as the command value to the actuator.

Here, H represents a time constant when the proxy converges to the target position. F is an upper limit value of the force generated by the linear actuator 403 as a moving mechanism, K is a proportional gain, B is a differential gain, p _s is a proxy position, p _d is a position command value, and p is a collision object 404. The position, that is, the encoder detection value, sgn (x) represents the sign function of the equation (4).

  The generated force g generated by the PSMC 701 is a force applied to the proxy from the sliding mode controller (SMC905), and is represented by g on the upper side of the equation (5). Further, a force (virtual viscoelastic force) applied to the proxy from the virtual connecting element 904 (see FIG. 9) is represented by a lower g in the equation (5), and an upper g in the equation (5). Is the same power. Note that details of the derivation process of Equation (5) are described in Patent Document 1, for example.

<4-5. Impact test conditions>
Various conditions in the collision test will be described below.
<< 4-5-1. Admittance control parameters >>
The admittance control parameters are the following parameters for calculating the above-described equation (1) by the virtual object motion calculator 410 of the admittance controller 401. That is, the admittance control parameters are a virtual force (target force) f _{d applied to} the virtual object, a mass m _{d of} the virtual object, and a viscosity c _d of the virtual object. These f _d, m _d, and the following values in advance as c _d.
f _d : 5 [N]
m _d : 0.7 [kg]
c _d : 0.008 [Ns / mm]

<< 4-5-2. Position controller parameters >>
The position controller parameters are the following parameters for calculating the above-described equations (2) to (5) in the position controller 420 of the admittance controller 401. That is, the position controller parameters are the proportional gain K, the differential gain B, the upper limit value F of the force generated by the actuator, and the time constant H for the embodiment for the comparative example and the embodiment. The following values were preset as K, B, F, and H.
K: 1000 [N / mm]
B: 3 [Ns / mm]
F: 45 [N]
H: 0.1 [s] (used only for PSMC701)

<< 4-5-3. Other verification conditions >>
As other verification conditions, the following values were measured.
Real quantity m of impact body: 4.5 [kg]
Buffer material: EPDM sponge thickness t: 10 [mm]
Speed immediately before the collision: v = 150 [mm / s]
EPDM is ethylene-propylene-diene rubber.

<4-6. Test results>
A verification device (comparative example) when the output limiting PD controller 601 shown in FIG. 6 is adopted as the position controller 420, and a verification device (example) when the PSMC 701 shown in FIG. 7 is adopted as the position controller 420. Were used to perform a crash test under the above crash test conditions. In the collision test, for example, a collision body having a substantial amount of 4.5 kg is moved on a straight line, and is collided with the collision object at a speed of 0.54 km / hour immediately before the collision. This collision test assumes, for example, a case where a person accidentally collides with and puts a hand under a robot arm that moves linearly from top to bottom. Based on the result of the collision test, the verification device of the comparative example and the embodiment verifies whether the control that immediately generates the thrust in the direction in which the robot arm is separated from the person can be stabilized in order to reduce the impact on the person. It becomes possible.

As a verification result by each verification device, FIG. 8 shows a temporal variation of the contact force f measured when the collision body 404 collides with the collision object 406. The horizontal axis of the graph shown in FIG. 8 indicates time t [seconds], and the vertical axis indicates contact force f [N]. A line (thin solid line) with a contact force of 5 [N] indicates a target force f _d of a preset admittance control parameter. The broken line indicates the contact force f of the colliding body by force control (comparative example) when the base position control is output limited PD control. The thick solid line indicates the contact force f of the collision object by force control (example) when the position control of the base is PSMC.

The comparative example corresponds to the verification of the effect of a conventional control device using PD control for position control inside force control. Specifically, in the collision test, a mass (virtual object mass m _d = 0.7 kg) corresponding to about 15% of the actual mass with respect to the collision object (actual mass m = 4.5 kg) corresponding to the robot arm is virtually assumed. Set as mass.

In this example, as shown in the figure, the collision occurred when t = 0.25 seconds. In the case of the comparative example, as indicated by a broken line in FIG. 8, when t = 0.3 seconds (0.05 seconds after the collision), the contact force f slightly exceeds the target force f _d (about 10 N). . Thereafter, the contact force f becomes 0 N or a negative contact force. When t = 0.4 seconds (0.15 seconds after the collision), the contact force f barely falls within the upper limit F (= 45 N) of the force generated by the actuator. A contact force of 40 N was shown. Thereafter, the contact force f becomes 0 N or a negative contact force, but when t = 0.6 seconds (0.35 seconds after the collision), the upper limit value F (= 45 N) of the set force is far exceeded, A contact force of about 110 N, which is 22 times the target force f _d , was shown. Similarly, as shown by a broken line in FIG. 8, a waveform in which a large contact force and zero or negative contact force appear alternately in time appears six times per second after the collision, and one second or more has elapsed after the collision. However, the phenomenon that the contact force vibrates was observed.
In addition, although illustration is abbreviate | omitted, when the virtual mass was set more largely for stabilization (in order to eliminate a vibration), the contact force (collision force) became large. In addition, vibration was observed when the virtual mass was set smaller to reduce the contact force.

In the case of the example, as shown by a thick solid line in FIG. 8, when t = 0.3 seconds (0.05 seconds after the collision), the target force f _d is slightly exceeded (about 10 N) as in the comparative example. ) Shows the contact force f. Thereafter, the contact force did not increase any more, damped and oscillated toward the target force f _d , converged 0.2 to 0.3 seconds after the collision, and was stable. Since the embodiment uses a control device that uses PSMC for position control inside the force control under the same collision test conditions as the comparative example, the contact force is reduced and the vibration converges in a short time. It has been verified that the effect of reducing the instability of the admittance control can be reduced. Note that the impact reduction effect is an example, and the experiment conducted for the impact force verification is an example.

  It was difficult to assume in advance what kind of phenomenon would occur by replacing conventional PD control with PSMC in position control within force control. Therefore, as shown in the above-described test results, it is a remarkable effect that cannot easily be conceived that the vibration of the contact force is mitigated by using PSMC for position control inside the force control.

  As described above, the robot control apparatus 1 according to the embodiment of the present invention has a limit in the reduction of virtual mass set by conventional control using PD control or the like for position control that is a base of force control. Even so, destabilization of admittance control can be mitigated.

  As mentioned above, although preferable embodiment of the control apparatus of this invention was described, this invention is not limited to above-described embodiment. For example, the PSMC is not limited to the PD type, but may be a P type or a PID type.

  In the present embodiment, the control device has been described as performing the force control of the robot arm. However, the present invention is not limited to the force control of the robot arm, and can be widely applied to admittance control in general. In this case, an effect that contributes to stabilization of admittance control in general can be achieved.

DESCRIPTION OF SYMBOLS 100 Human-robot collaboration system 1 Robot control apparatus 2 Target force setting means 3 Force control means 4 Servo amplifier 5 Target force acquisition means 6 Contact force acquisition means 7 Virtual object motion calculation means 8 Position control means 10 Arm 11 Force sensor 12 Actuator DESCRIPTION OF SYMBOLS 14 Servo motor 15 Encoder 20 Cable 401 Admittance controller 402 Servo amplifier 403 Linear motion actuator 404 Colliding body 405 Force sensor 406 Impacted body 407 Buffer material 410 Virtual object motion calculator 420 Position controller 501 Virtual object 502 Contacted object 503 Force generation element 504 Virtual damper 601 Output limited PD controller 602 Control target 603 Unit saturation function 701 PSMC
702 Control object 703 Sign function 901 Target position 902 Control object 903 Proxy 904 Virtual holding element 905 Sliding mode control P Person R Robot

Claims (4)

Contact is generated between an external object and another object by including force control means having a position control means inside, and the force control means performing admittance control based on force information input from the outside. A control device for controlling force,
The position control means uses proxy-based sliding mode control. The position control means includes
A feedback controller including a proportional control for performing feedback control including at least a proportional action;
A sliding mode controller based on the sliding mode control law,
The proxy-based sliding mode control is performed by cooperating a feedback controller including the proportional control and the sliding mode controller,
The feedback controller including the proportional control is added to the proxy from a virtual connection element that virtually connects a proxy that is a point-like virtual object having no mass in a virtual space and the external object. Calculated virtual viscoelastic force
The sliding mode controller is a force that is balanced with the viscoelastic force and is a virtual force applied to the proxy when performing control to converge the position of the proxy to a target position set in the sliding mode controller. As a thrust applied to the object,
The control device according to claim 1. The object includes a drive mechanism that generates thrust applied to the object and a position detection sensor,
The force control means includes
A virtual mass, a virtual viscoelasticity, and a virtual added to the virtual object are predetermined for a virtual object having a predetermined non-zero mass that is less than a substantial amount of the object corresponding to the object. Virtual object motion calculation means for calculating a target position of the virtual object by solving a motion equation for the virtual object including each parameter of the target force as force using the contact force as an input,
The position control means receives the calculated target position and the position of the object detected by the position detection sensor as input, and provides torque (thrust) to a drive mechanism that drives the object using the proxy-based sliding mode control. The control device according to claim 1 or 2, wherein a command value is output. Contact is generated between an external object and another object by including force control means having a position control means inside, and the force control means performing admittance control based on force information input from the outside. A control method by a control device for controlling force,
The position control means uses proxy-based sliding mode control.

Images