JP5652906B2 - Robot controller - Google Patents

Description

  The present invention relates to a robot control apparatus that switches a continuous operation of a robot.

  2. Description of the Related Art A horizontal articulated transfer robot that transfers an object to be transferred such as a semiconductor wafer in a narrow place such as a vacuum chamber is known (see, for example, Patent Document 1). Such a transport robot is configured to transport a transported object by causing a hand connected via a plurality of arms to move straight (expand / contract) or rotate (turn). For each operation, an optimal controller is set to perform high-speed and high-accuracy conveyance (see, for example, Patent Document 2).

Japanese Patent Laid-Open No. 10-329069 JP 2003-340755 A

Masakazu Mukai, “Robust Model Predictive Control Using Invariant Sets for Hybrid Systems”, Doctoral Dissertation, Department of Functional Development Science, Graduate School of Natural Sciences, Kanazawa University, 2005 Kenji Hirata, “Analysis of constraint systems using invariant sets and application to control system design”, System / Control / Information: Journal of System Control Information Society, Vol. 47, no. 11, pp. 507-513, 2003

  When a semiconductor wafer is transferred by such a transfer robot, as shown in FIG. 13 (a), a controller set for the straight movement operation (extension operation and contraction operation) and a rotation operation are set. The controller is switched (switched). Conventionally, as shown in the portions surrounded by the thick lines in FIGS. 13B and 13C, a stop time is provided when switching the respective controllers. That is, since a stable stop time is required so that the hand trajectory does not collapse (see FIG. 14), it has not been possible to smoothly shift from a straight traveling operation to a rotating operation. As a result, there has been a problem that the transport throughput is reduced.

  In this regard, when performing controller switching control, it is possible to prevent a decrease in throughput by applying model predictive control using an invariant set to the transfer robot. 2 is suggested. However, Non-Patent Document 1 and Non-Patent Document 2 specifically describe how to design the controller switching timing and the initial state of the controller after switching, for example, when shifting from a straight running operation to a rotating operation. It is not disclosed.

  An object of the present invention is to provide a robot control device that allows a robot to smoothly transition from one operation to the other.

  The problem to be solved by the present invention is as described above. Next, means for solving the problem will be described.

The robot control device according to claim 1 , wherein the robot control device switches a continuous operation of the robot including a hand that conveys the object to be conveyed and a plurality of motors that drive the hand, and corresponds to the first operation of the robot. And a second controller corresponding to a second operation of the robot, wherein the state of the robot defined by a rotation angle and an angular velocity of the plurality of motors is the first operation. At the time of switching from the second operation to the second operation, if the hand shake belongs to an invariant set derived from a constraint condition that is less than a specified value , the initial state of the second controller corresponding to the second operation is It is calculated and dynamically switched from the first controller to the second controller .

According to a second aspect of the present invention, the robot is a transfer robot that enables a straight movement operation for moving the hand straight and a rotation movement for rotating the hand, and the first operation is either the straight movement operation or the rotation operation. However, the second operation is the other of the rectilinear operation and the rotation operation.

According to a third aspect of the present invention, the robot control device derives a low-dimensional invariant set by projecting a multi-dimensional invariant set defined by the constraint condition onto the state space of the transfer robot.

In claim 4, the initial state of the second controller corresponding to said second operation, when the switching to the second operation from the first operation, the shake of the hand is less than a specified value It is calculated from an initial state determination method using a linear equation that satisfies a constraint condition.

In claim 5, the initial state of the second controller corresponding to the second operation is set to minimize the fluctuation of the control input to the transfer robot when switching from the first operation to the second operation. Is calculated from the initial state determination method.

  As effects of the present invention, the following effects can be obtained.

In the first aspect, the robot can smoothly move from the first operation to the second operation without stopping.

  According to the second aspect of the present invention, the transfer robot can smoothly shift from the straight operation to the rotation operation or from the rotation operation to the straight operation without stopping. Therefore, the conveyance throughput can be improved.

  According to the third aspect of the present invention, the switching determination between the two controllers can be performed only by the state of the transfer robot.

  According to the fourth aspect of the present invention, it is easy to calculate the initial state of the controller.

  According to the fifth aspect of the present invention, the direct control input to the transfer robot is unlikely to become discontinuous when switching between the two controllers.

The figure which shows a wafer conveyance robot and a chamber. The figure which shows the structure of a wafer conveyance robot. (A) The figure which shows the straight-ahead operation | movement of a wafer transfer robot. (B) The figure which shows rotation operation | movement of a wafer transfer robot. Conditional feedback structure combined with gain scheduling. The conceptual diagram of the linearization with respect to a nonlinear system. Conceptual diagram of invariant set. The conceptual diagram showing the relationship of each invariant set. Sectional view of the O _p with respect to the rotational operation of the wafer transport robot. The flowchart which shows the control aspect of the wafer conveyance robot to which the switching control law which concerns on this invention is applied. (A) The figure which shows the locus | trajectory of the conventional hand. (B) The figure which shows the locus | trajectory of the hand of this invention. The figure which shows the experimental result of the switching control law using an invariant set. The figure which shows the other experimental result of the switching control law using an invariant set. (A) The figure which shows the existing switching control system. (B) The figure which shows the timing chart of the existing switching control. (C) The figure which shows the timing chart of the existing switching control. The figure which shows a mode that the locus | trajectory of a hand collapses by switching of operation | movement.

  Next, the configuration of the wafer transfer robot 1 according to an embodiment of the present invention will be described.

  As shown in FIG. 1, the wafer transfer robot 1 loads and unloads a silicon wafer 60 having a diameter of about 300 mm into a processing chamber called a chamber 50 surrounding the wafer transfer robot 1. The chamber 50 includes, for example, a load lock chamber 51 that carries the silicon wafer 60 into and out of the chamber, a process chamber 52 that performs processing such as CVD (Chemical Vapor Deposition), etching, and the like.

  As shown in FIG. 2, the wafer transfer robot 1 is a horizontal articulated robot, and includes a hand 2, two upper arms 3 and 3, a lower arm 4, a link member 5, a first motor 10, The second motor 20, the first servo amplifier 15, the second servo amplifier 25, and the robot control device 30 are mainly configured.

  Both ends 2a and 2a of the hand 2 are formed in a fork shape, and a silicon wafer 60 is placed thereon.

  The upper arms 3 and 3 are each formed in a substantially “<” shape in plan view, and the bent portions thereof are arranged to face each other, and the respective ends 3 a and 3 a are slightly inside the both ends 2 a and 2 a of the hand 2. It is linked and linked.

  The lower arm 4 is formed in a substantially elliptical shape in a plan view, and is disposed below the upper arms 3, 3. In the interior thereof, there are two pulleys 6, 6, and the two pulleys 6, 6. The wound first belt 7 is accommodated.

  The link member 5 is a plate material having a substantially triangular shape in plan view, and is provided between the upper arms 3 and 3 and the lower arm 4. The link member 5 has a hole 5a, and two rotating shafts 5b and 5b project from the upper surface thereof. A rotating shaft 8 is inserted and fixed in the hole 5a and the shaft hole of one (tip side) pulley 6 in the lower arm 4. Pulleys 11 and 11 are pivotally supported on the rotary shafts 5b and 5b, respectively. One end of the second belt 9 is fixed to the pulleys 11 and 11, and the middle portion of the second belt 9 is wound around the rotary shaft 8.

  The first motor 10 is disposed at the lower part of the other arm (base side) of the lower arm 4, and its output shaft is fixed to the lower arm 4.

  The second motor 20 is disposed below the first motor 10, and its output shaft is fixed to the other pulley 6 in the lower arm 4.

The first servo amplifier 15 is connected to the first motor 10 and the robot control device 30, and supplies power to the first motor 10 based on a signal from the robot control device 30 to rotate the first motor 10. The first servo amplifier 15, the rotation angle theta _a and the angular velocity d [theta] _{a /} dt of the output shaft of the first motor 10 is configured to detect the rotation angle theta _a and the angular velocity d [theta] _{a /} dt detected, the robot control Transmitted to the device 30.

The second servo amplifier 25 is connected to the second motor 20 and the robot control device 30, and supplies power to the second motor 20 based on a signal from the robot control device 30 to rotate it. The second servo amplifier 25 is configured to be able to detect the rotation angle θ _b and the angular velocity dθ _b / dt of the output shaft of the second motor 20, and the detected rotation angle θ _b and angular velocity dθ _b / dt are controlled by the robot. Transmitted to the device 30.

The robot control device 30 calculates torques τ _a and τ _b that serve as control inputs for the first motor 10 and the second motor 20 based on the rotation angles θ _a and θ _b and the angular velocities dθ _a / dt and dθ _b dt. Thus, signals corresponding to the torques τ _a and τ _b are transmitted to the servo amplifiers 15 and 25.

  The wafer transfer robot 1 can perform a straight movement operation (extension operation and contraction operation) for moving the hand 2 straight as shown in FIG. 3A and a rotation operation for rotating the hand 2 as shown in FIG. Composed. These two operations are realized when the robot control device 30 gives control targets (target rotation angles) as shown in Table 1 to the first motor 10 and the second motor 20.

That is, when the first motor 10 and the second motor 20 are rotationally driven simultaneously (both θ _a and θ _b are increased or decreased), the wafer transfer robot 1 is rotated. At this time, the target rotation angle of each motor is set so that | θ _a −θ _b | is as small as possible. When the first motor 10 is driven to rotate while the second motor 20 is driven so as to maintain the rotation angle θ _b (θ _a is increased or decreased while holding θ _b ), the wafer transfer robot 1 , It goes straight. At this time, | θ _b | affects the accuracy of the transport track. In this way, the silicon wafer 60 is transferred by appropriately switching between the two operations.

The robot control device 30 switches the continuous operation of the wafer transfer robot 1. The continuous operation of the wafer transfer robot 1 means one different operation and the other operation that the wafer transfer robot 1 performs continuously.
The robot controller 30 includes a controller C ₁ corresponding to the straight operation, a controller C ₂ corresponding to the rotation operation, the. The controller C ₁ and the controller C _2, which corresponds to the operations respectively continuous. The operation of the wafer transfer robot 1 corresponding to the controllers C ₁ and C ₂ is performed by the robot controller 30 switching between the controllers C ₁ and C ₂ . The dynamic switching rule of the controller for such a control system having a plurality of operation modes is shown below.

Specifically, a description will be given assuming a control system of the wafer transfer robot 1 that rotates from the position where the hand 2 (specifically, the arm of the hand 2) extends.
(1) The robot is at an initial set point (the hand 2 is extended).
(2) Applying the controller C ₁ for the straight movement to the extended hand 2 to start the shrinking movement, which is one of the straight movements.
(3) contraction operation in unfinished state, by applying the controller C ₂ of rotation, starts to rotate.
When the hand 2 rotates, it is required that the shake of the hand 2 is always equal to or less than a specified value, which is the same when the controller is dynamically switched. Therefore, an invariant set that guarantees that the shake of the hand 2 is equal to or less than a specified value during the rotation operation of (3) is defined, and a dynamic switching rule of the controller using this is considered.

First, in order to design the controllers C ₁ and C ₂ , the wafer transfer robot 1 to be controlled is modeled.

A non-linear model of the wafer transfer robot 1 is expressed as shown in Equation 1 using a general robot equation of motion. M (θ) represents an inertia matrix such as moment of inertia, h (θ, dθ / dt) represents a non-linear term such as centrifugal force or Coriolis force, and θ and τ are represented as in Equation 2. Here, θ _a and θ _b represent rotation angles [rad] of the first motor 10 and the second motor 20, and τ _a and τ _b are control inputs to the first motor 10 and the second motor 20. Represents torque [Nm].

In addition, when Formula 1 and Formula 2 are expressed more generally, they can be expressed as Formula 3 and Formula 4. _xp represents the state of the wafer transfer robot 1, and u represents a control input to the wafer transfer robot 1.

Then, it is called generally state equations plus the equation for the controlled variable z ₁ and observables y in Equation 4 (see Equation 5).

  When designing a controller for a nonlinear system as shown in Equation 5, a widely used method is a method of designing a controller by performing linearization near the equilibrium point. Specifically, a linear controller for a linear system obtained by applying Taylor's theorem to Equation 5 and approximating so as to ignore higher-order terms, assuming a small operation at the equilibrium point. To design.

It is known that the equilibrium point of the equation of state shown in Equation 5 is expressed as Equation 6 using an arbitrary angle. This means that the wafer transfer robot 1 is stationary at an arbitrary posture without any control input u being given. In the following formulas, values with superscript bars (x _p , u, θ, r, y, etc.) represent respective values at the equilibrium point.

  In the contraction operation and the rotation operation, considering Equation 7 as this arbitrary angle, the equilibrium point of the contraction operation corresponds to the posture in which the wafer transfer robot 1 extends the arms (upper arm 3 and lower arm 4) most. The equilibrium point of the rotational motion corresponds to the posture in which the arm is folded.

  From Equations 2, 6, and 7, the equilibrium point of the state of the wafer transfer robot 1 in each operation is expressed by Equation 8.

Here, the target value follow-up control of the controlled amount z ₁ is considered. The external setpoint signal should follow the controlled variable z _1, represented by r.

Assuming that there is a leveling point for the shrinking operation and the corresponding control inputs (superscript bar x _p , ₁ and superscript bar u ₁ ), Equation 5 is expressed by Equation 9.

Assuming that there is a neutral point of rotation and corresponding control inputs (superscript bar x _p , ₂ and superscript bar u ₂ ), Equation 5 is expressed as Equation 10.

The controllers C ₁ and C ₂ for the respective operations are respectively designed for the following linear state equations around these equilibrium points (superscript bar x _p , ₁ and superscript bar x _p , ₂ ). . First, the contraction operation is expressed by Equations 11 and 12, and the rotation operation is expressed by Equations 13 and 14.

Next, a controller designed for the wafer transfer robot 1 linearized in this way is expressed by Equations 15 and 16. Equation 15 is a controller C ₁ designed for straight-ahead operation, and Equation 16 is a controller C ₂ designed for a rotational operation.

Next, the first closed loop system composed of the wafer transfer robot 1 (the hand 2 is extended) and the controller C ₁ is expressed by Equation 17, and the wafer transfer robot 1 (the hand 2 is folded) and the controller A second closed-loop system composed of C ₂ is expressed by Equation 18.

Next, a design is made for the control system of the straight running operation. That is, the design controller _{C 1.} Assuming that the control input using the inertia matrix as in Expression 19 is τ with respect to Expression 1, Expression 1 is expressed by Expression 20. Here, α = θ _a −θ _b .

When Equation 20 is linearized at the equilibrium point of x _p , ₁ , it is approximated as Equation 21.

This indicates that the plant in question is completely incoherent with linearization and does not depend on θ. A two-degree-of-freedom control system with a conditional feedback structure based on gain scheduling as shown in FIG. 4 was designed for the number 21 of linearly incoherent plants. W (θ) is a function that depends only on θ, and here W (θ) = M (θ). T _ref is a plant model expression, and s is a Laplace operator. F designates a target value response, K designates a feedback characteristic, and R is feedforward control. The parameters of each element such as feedback gain and phase compensation in the controller C ₁ designed in the wafer transfer robot 1 are expressed by Equation 22.

Next, the control system for the rotational operation is designed. That is, the design controller _{C 2.} Formula 1 is expanded to Taylor. With the equilibrium point as the origin posture, superscript bar θ _a = superscript bar θ _b = 0, and the obtained linear state equation is expressed by Equation 23.

Superscript bar α = superscript bar θ _a −superscript bar θ _b . Here, since the inertia matrix depends on the rotation angle difference α = θ _a −θ _b , the superscript bar α = 0 can be set. That is, Expression 23 is expressed by Expression 24.

For a linearized plant of Formula 24, a two-degree-of-freedom control system with a conditional feedback structure based on gain scheduling was designed in the same manner as a control system for linear operation. The parameters of each element such as feedback gain and phase compensation in the controller C ₂ designed in the wafer transfer robot 1 are expressed by Equation 25.

  As shown in Equations 22 to 25, an optimal control system is designed for each operation in the straight operation and the rotation operation.

  Next, a switching rule between two controllers that is a main part of the present invention will be described.

  First, the problem of switching control is generalized. As shown in FIG. 5, when two linearized systems are defined, it is assumed that the first system is switched to the second system. The first system corresponds to the first closed-loop system shown in Expression 17, and the second system corresponds to the second closed-loop system shown in Expression 18.

Here, in order to derive the ideal switching rule shown in FIG. 10B, an algorithm using an invariant set is proposed.
Next, an invariant set corresponding to the constraint condition is constructed. Invariant sets are known as a useful tool for designing in consideration of system constraints, and various methods have been proposed and their effects have been demonstrated. As shown in FIG. 6, the invariant set is characterized in that a state that has entered the set cannot be removed from the set thereafter. To construct an invariant set, first, the amount of restraint is defined.

When this is applied to the wafer transfer robot 1, as shown in Table 1, | θ _a −θ _b | becomes small with respect to the rotation operation that is the operation after switching that constitutes the second closed loop system. Must be controlled. Therefore, the restrained amount z ₀ is set as shown in Equation 26.

Number 26, using the x _p defined by the number 3, can be written as the number 27.

For example, when the constrained amount z ₀ is set to be 0.1 or less, it can be expressed as Equation 28. Further, when Expression 28 is generally expressed, Expression 29 is obtained.

Since Equation 29 is for the nonlinear plant of Equation 5, it can be expressed as Equation 30 if these are further based on the equilibrium point (superscript bar x _p , ₂ ) of the rotational operation.

The constraint condition can be expressed strictly as Equation 31. Here, as shown in Equation 8, since superscript bar x _p , ₂ = 0, it is noted that z ₀ [t] = z ₀ , ₂ [t].

Then, the number 31 as a set _{Z 2,} expressed as in equation 32. M _z and m _z are parameters that define the constraint conditions.

Here, when the external input w is given to the second closed-loop system, the invariant set O _∞ , ₂ for achieving the constraint of Z ₂ in Expression 25 is defined as Expression 33. Equation 33 is obtained as a convex polyhedron. Z ⁺ represents a set of positive integers.

Project the invariant set O _∞ , ₂ onto the state space of the linear model of the plant. The projected invariant set is defined as O _p , ₂ and defined as Equation 34.

Furthermore, the invariant set O _p , ₂ is translated into the state space of the nonlinear plant. The translated invariant set is defined as O _p and defined as Equation 35.

FIG. 7 is a diagram showing the relationship between the invariant sets O _∞ , ₂ · O _p , ₂ · O _{p in} a geometrically easy-to-understand manner. However, the dimensions of x _p , ₂ and x _c , ₂ are 1 in the figure. Referring to FIG. 7, first, a coordinate system (x _p , ₂ , x _c , ₂ ) formed by a linearized plant and a controller is defined on the state space of the nonlinear plant. Next, an invariant set O _∞ , ₂ in the region indicated by the thick solid line is constructed. Subsequently, the invariant set O _∞ , ₂ is projected onto the linearized plant state space to obtain the invariant set O _p , ₂ . Then, the invariant set O _{p, 2} moves in parallel, invariant set O _p in the burst to the state space of the nonlinear plant is obtained.

Here, it is assumed that a certain state x _p of the nonlinear plant is given by a point inside the invariant set O _p as shown in FIG. Then, it is also _{x p, 2} corresponding to _{x p} in invariant set _{O p, 2,} state _{x c, 2} corresponding controller are not unique. Therefore, it is necessary to determine x _c , ₂ (initial state x _c , ₂ ) after switching determination by some method. This determination method is currently based on the condition shown in Equation 36.

The outline of the switching rule algorithm using invariant sets is shown below.
(1) Drive in the first closed loop system.
(2) Judge whether to enter the invariant set.
(3) If (2) is satisfied, immediately switch to the second closed loop system and start driving.
A switching rule algorithm that satisfies the procedures (1) to (3) is constructed. Thereby, after switching, the second closed-loop system is guaranteed to be constrained. However, since the state x _c , ₂ of the controller C ₂ in the second closed loop system is unknown, it is necessary to determine an appropriate state x _c , ₂ .

Based on the above procedure, the obtained invariant set _Op is shown in FIG. However, the invariant set _Op is represented by Equation 37 as the rotation angle difference | θ _a −θ _b | between the first motor 10 and the second motor 20 so that the value c is 0.02 or less. Set to. In addition, the projection algorithm of Fourier Motzkin was used for the calculation of the projection. It should be noted that the invariant set _Op is originally in the four-dimensional space because the order of the plant is the fourth order, and it is difficult to illustrate the set. However, in before the switching straight operation, it controls the second motor 20 such that a change in the rotation angle theta _b holds lest the angle theta _b occur. FIG. 8 takes this into consideration and shows a diagram in which θ _b = 0 and dθ _b / dt = 0.

Here, at time _{t s,} and x _{p [t s]} ∈O p has been achieved. At this time, the second closed-loop system the initial state of the corresponding controller _{C 2} to _{x c, 2 [t s]} ∈O ∞ , as described above, ₂ is not unique. Therefore, the initial state x _c , ₂ is determined as follows.

The invariant set O _∞ , _p is divided so as to correspond to the state x _c , ₂ of the wafer transfer robot 1 and the state x _p , ₂ of the controller, as expressed in Equation 38.

What is necessary is just to find x _p , ₂ that satisfies the equation 36, which is an inequality. Therefore, Equation 38 is regarded as an equation and developed as Equation 39.

Here, if M _c , ₂ is regular, x _p , ₂ can be uniquely determined by an inverse matrix, but generally M _c , ₂ is not a square matrix. Therefore, x _p , ₂ is obtained from the linear equation of Formula 40 assuming a pseudo inverse matrix M _c , ₂ ^† . Thus, the determined x _p , ₂ is determined as the initial state x _c , ₂ .

As another method of determining the initial state x _c , ₂ , there is a method of solving a linear programming problem so that the fluctuation of the control input before and after the switching time is minimized. This is based on simulations and experimental results that the direct control input to the wafer transfer robot 1 tends to be discontinuous at the time of switching, and is used to avoid this.

That is, before time _t s -1 during switching determination, determines that the input _u are actually applied to the control system _[t s -1], the time _t the state in _{s x c,} 2 a _{[t s]} Is the method. When this is described as the form of the optimization problem, the following equation 41 is obtained.

  As shown in Formula 41, by evaluating 1 norm, the optimization problem is reduced to a linear programming problem.

  A control mode in which the operation switching rule as described above is applied to the wafer transfer robot 1 to switch from the shrinking operation to the rotating operation will be described with reference to FIG.

In step S1, the wafer transfer robot 1 is driven by a contraction operation. That is, driven by the first closed-loop system consisting of the wafer transport robot 1 and the controller C _1. Thus, the robot controller 30, the rotational angle theta _b of the rotation angle theta _a and the second motor 20 of the first motor 10 drives and controls the first motor 10 and second motor 20 to a predetermined angle The hand 2 is contracted. Then, the process proceeds to step 2.

In step S2, the robot controller 30 observes the state _{x p} of the wafer transfer robot 1. That is, the robot control device 30 detects the rotation angle θ _a of the first motor 10, the angular velocity dθ _a / dt, the rotation angle θ _b of the second motor 20, and the angular velocity dθ _b / dt. Then, the process proceeds to step S3.

In step S3, the robot controller 30 determines whether x _p εO _p . That is, the robot controller 30, _a theta was observed in step _{2, dθ a / dt, θ} b and d [theta] _b / dt determines whether belonging to the invariant set _{O p} calculated from the number 35 expression . If it belongs, the process proceeds to step S4. When not belonging, it transfers to step S2 again.

In step S4, it determines the initial state _{x c, 2} controller _{C 2.} In other words, the robot control device 30 determines the initial state x _c , ₂ of the controller C _{2 using} Equation 37 or Equation 38. Then, the process proceeds to step 5.

In step S5, the wafer transfer robot 1 is driven by a rotation operation. That is, in the robot control device 30 is switched from the controller C ₁ to the controller C _2, driven by a second closed loop system composed of the wafer transport robot 1 and the controller C _2. Thus, the robot controller 30, the rotational angle theta _b of the rotation angle theta _a and the second motor 20 of the first motor 10 drives and controls the first motor 10 and second motor 20 to a predetermined angle The hand 2 is rotated. In this way, switching from the shrinking operation as one operation to the rotating operation as the other operation is performed.

  FIG. 10A shows the trajectory of the conventional hand, and FIG. 10B shows the trajectory of the hand 2 of the present invention. Comparing the conventional trajectory with the trajectory of the present invention in the area surrounded by the dotted line, it can be seen that the trajectory of the present invention smoothly shifts from the contraction operation to the rotation operation and becomes an ideal trajectory. Further, as shown in the two-dot chain line in FIG. 10B and FIG. 14, the trajectory of the hand 2 is not broken.

  FIG. 11 shows the experimental results when the initial state determination method for solving the linear equation is used, and FIG. 12 shows the experimental results when the initial state determination method that minimizes the change in the control input is used. Both operations are switched in 14 seconds. When FIGS. 11A and 11B are compared with FIGS. 13A and 13B, it can be seen that the transport throughput in the entire operation is improved. Similarly, when FIGS. 12A and 12B are compared with FIGS. 13A and 13B, it can be seen that the overall throughput is improved.

Further, as shown in FIG. 11 (e) and FIG. 12 (e), the angular difference between the rotational angle theta _a rotation angle theta _b it can be seen that is suppressed to 0.02.

Further, in the case where FIGS. 11C and 11D are compared with FIGS. 12C and 12D, the case of using the initial state determination method that minimizes the change in the control input is used. It can be seen that the control input, that is, the torque τ _a · τ _b is small.

  Note that the operation switching using the invariant set can be similarly applied not only to the contraction operation to the rotation operation but also to the rotation operation to the extension operation. Further, the present invention is not limited to the wafer transfer robot 1 and can be applied to various robots.

Note that the wafer transfer robot 1 may be configured to be capable of performing an operation of moving the hand 2 straight while rotating the hand 2, that is, a rectilinear operation involving a rotating operation. In this case, the robot controller 30 further includes a controller C ₃ corresponding to the straight operation with rotation. Migration Thus, between the controller _C 1 _{· C} 2 _{· C 3,} by switching with the invariant set _{O p,} smoothly from one operation corresponding to these controllers _C 1 _{· C} 2 _{· C 3} to the other operation can do. In other words, straight operation from rotation operation, rotation operation to straight operation, straight operation from rotation to straight operation, straight operation with rotation operation, straight operation from rotation operation, straight operation from rotation operation, straight operation with rotation operation, straight operation with rotation operation It is possible to make a smooth transition from rotating to rotating.

Note that a three-axis wafer transfer robot having a lift axis is configured to be able to lift and lower the hand 2 up and down. In this case, the robot control device includes a controller C ₄ corresponding to the vertical movement. Further, the three-axis wafer transfer robot has an operation of rotating the hand 2 while moving it up and down, that is, a rotating operation accompanied by an elevating operation, and an operation of moving the hand 2 straight up and down, that is, a linear operation accompanied by an elevating operation. May be possible. In this case, further comprises a controller C ₅ corresponding to the rotation operation with lifting operation, the controller C ₆ corresponding to the straight operation with vertical movement, the. Thus, the controller _C between _{1 · C 2 · C 3 ·} C 4 · C 5 · C 6, by switching with the invariant set _{O p,} these controllers _{C 1 · C 2 · C 3} · C 4 · C It is possible to smoothly shift from one operation corresponding to ₅ · C ₆ to the other operation. That is, it is possible to smoothly shift from a lifting operation to a rotating operation, from a rotating operation to a lifting operation, from a lifting operation to a rectilinear operation, from a rectilinear operation to a lifting operation, from a rotating operation involving a lifting operation to a rectilinear operation, etc.

  In the chamber 50, when many process chambers 51 surrounding the wafer transfer robot 1 are provided, or when the load lock chamber 52 is provided in multiple stages, the number of times the silicon wafer 60 is carried into and out of the process chamber 51 is large. Become. Accordingly, the ratio of the transfer time for the wafer transfer robot 1 to transfer the silicon wafer 60 to the processing time of the entire processing apparatus increases. According to the present invention, since the air cut time of the wafer transfer robot 1 itself, that is, the transfer time can be shortened, the throughput of the entire processing apparatus can be improved.

  As described above, the robot control device 30 that switches the continuous motion of the robot, includes a plurality of controllers corresponding to the motion, and the state of the robot is an invariant set derived from the constraint condition in the motion after the switching. Belongs to the controller, the initial state of the controller corresponding to the operation after switching is calculated, and the controller is dynamically switched. As a result, the robot can smoothly transition from one operation to the other without stopping.

In addition, the robot is a wafer transfer robot 1 that includes a hand 2 that transfers a silicon wafer 60 that is a transfer object, and that is capable of performing a rectilinear operation for moving the hand 2 straight and a rotating operation for rotating the hand 2. The controllers C ₁ and C ₂ are respectively provided corresponding to the linear movement operation and the rotation operation. As a result, the wafer transfer robot 1 can move smoothly from the straight operation to the rotation operation or from the rotation operation to the straight operation without stopping. Therefore, the conveyance throughput can be improved. Further, the hand 2 is not shaken and collision with the chamber 50 can be avoided, so that the transfer can be performed with high accuracy.

Further, the invariant set O _p is the invariant set of multidimensional defined in constraint O∞, _2, to switch determines the operation derived by projecting into the state space of the wafer transfer robot 1 Is. Thus, it is possible to switch the determination of two operations only when x _p of the wafer transfer robot 1.

Further, the initial state x _c , ₂ of the controller C ₂ corresponding to the operation after switching is calculated from the initial state determination method based on the linear equation shown in Formula 40 that satisfies the constraint condition. Thereby, the calculation of the initial state x _c , ₂ of the controller C ₂ is facilitated.

Further, the initial state x _c , ₂ of the controller C ₂ corresponding to the operation after switching is calculated from an initial state determination method that minimizes the fluctuation of the control input to the wafer transfer robot 1 before and after switching. Thereby, at the time of switching, the direct control input to the wafer transfer robot 1 becomes difficult to be discontinuous.

1 Wafer transfer robot (transfer robot)
2 hand 3 upper arm 4 lower arm 5 link member 10 first motor 15 first servo amplifier 20 second motor 25 second servo amplifier 30 robot controller 50 chamber 60 silicon wafer (conveyed object)

Claims (5)

A robot control device that switches a continuous operation of a robot comprising a hand that transports a transported object and a plurality of motors that drive the hand ,
A first controller corresponding to the first movement of the robot, and a second controller corresponding to the second movement of the robot,
The state of the robot, which is defined by the rotation angles and angular velocities of the plurality of motors, is derived from a constraint condition in which the shake of the hand is equal to or less than a specified value when switching from the first operation to the second operation. A robot control that calculates an initial state of the second controller corresponding to the second operation and dynamically switches from the first controller to the second controller apparatus. The robot is a transfer robot that enables a straight-ahead operation for moving the hand straight and a rotation operation for rotating the hand,
2. The first operation according to claim 1, wherein the first operation is one of the rectilinear operation or the rotation operation, and the second operation is either the rectilinear operation or the rotation operation. The robot control device described. The robot control apparatus derives a low-dimensional invariant set by projecting a multi-dimensional invariant set defined by the constraint condition onto a state space of the transfer robot. The robot control apparatus according to 2. A linear equation that satisfies a constraint condition that the shake of the hand is equal to or less than a predetermined value when the initial state of the second controller corresponding to the second operation is switched from the first operation to the second operation. The robot control apparatus according to claim 1, wherein the robot control apparatus is calculated from an initial state determination method according to claim 1. An initial state determination method for minimizing fluctuations in control input to the transfer robot when the initial state of the second controller corresponding to the second operation is switched from the first operation to the second operation. The robot control device according to claim 1, wherein the robot control device is calculated from:

Images