JPH01217601A - Servo controller - Google Patents

Abstract

PURPOSE:To realize the drive of a servomotor with a minimum energy by performing the state feedback control with use of the real time of the servomotor and based on an optimum manipulated variable series and an optimum state track produced by a learning system. CONSTITUTION:An optimum manipulated variable series and an optimum state track can be obtained in response to a controlled system by performing a trial control action according to a control model decided previously by a learning system and repeating the trial actions while applying the self-correction to the deviation produced between the control model and an actual state. Then the state feedback control using the real time of a servomotor is carried out by a closed loop servo control system after the learning is through based on said optimum manipulated variable series and optimum state track produced by the learning system. As a result, the state control is possible in response to a model with high accuracy. Thus it is possible to reduce the overshoot that is caused by the linear feedback control and to accelerate the control time. Then the servomotor is driven with a minimum energy.

Description

Detailed Description of the Invention (Field of Industrial Application) The present invention relates to a digital servo control device that controls the position and speed of a mechanical system that is moved by a servo motor, such as a manipulator or a robot arm. In particular, the present invention relates to a servo control device capable of high-speed and highly accurate positioning of a servo motor, and high-speed start-up and sudden stopping during constant speed control and constant torque control.

(Prior Art) A digital servo control device using a microcomputer or the like is known as a control device that controls the position and speed of a mechanical system driven by a servo motor, such as a manipulator or a robot arm. The servo control system in this conventional digital servo control device basically consists of only a linear feedback system, most of which is dominated by PID control using an output feedback system.

(Problem M to be solved by the invention) However, with only such a digital linear feedback system, overshoot occurs during transient response. ! ! The disadvantage is that it greatly affects the fixed time.

In addition, since it is necessary to operate in the linear operation region, the feedback gain must be low, which results in a configuration that is vulnerable to external disturbances, and the stability of control operation is affected, especially when the amplitude limit of the manipulated variable cannot be ignored. warranty will no longer be possible.

Therefore, in a system that requires rapid dissolution, a design that takes into account the amplitude limit of the manipulated variable is required, which increases costs.

The present invention has been made in view of the above circumstances, and improves the response characteristics of a servo control system during transient response through repeated learning, eliminates as much as possible the overshoot that occurs in linear feedback control, and speeds up the settling time. An object of the present invention is to provide a servo control device that can drive a servo motor with minimum energy during a specified settling time given from design specifications.

(Means for Solving the Problems) In order to achieve the above object, the present invention provides a digital servo control device that controls the position and speed of a mechanical system driven by a motor, in which the mechanical system driven by the motor is constantly controlled. A means for generating an optimal manipulated variable sequence and an optimal state trajectory by repeatedly trying control based on the state deviation at the final step without a real-time feedback loop, and a method determined in advance to reduce the number of trials described above. A closed loop that performs real-time state feedback control of the servo motor based on the optimal state trajectory and optimal manipulated variable series generated by the learning system. It is characterized by having a servo control system.

(Function) In the servo control device according to the present invention, the learning system performs a control operation according to a predetermined control model, and repeats the trial operation while self-correcting the deviation between the model and the actual state. It is possible to generate an optimal manipulated variable sequence and optimal state trajectory corresponding to the object. In addition, after the learning is completed, the closed-loop servo control system performs real-time state feedback control of the servo motor based on the optimal state trajectory and optimal manipulated variable series generated by the learning system, so that the model can be accurately applied. Corresponding state control becomes possible.

(Example) .

Hereinafter, the present invention will be explained in detail based on illustrated embodiments.

FIG. 1 is a schematic configuration diagram of a servo control device showing an embodiment of the present invention. It consists of a closed-loop control system that controls the motor based on the optimal control conditions generated by the learning system.

Each component of the servo control device will be explained below with reference to FIG.

In Fig. 1, reference numeral 3 indicates a rewritable function generator (RAMI) for storing the optimum manipulated variable series.
, 8 is a rewritable function generator (RAM2) for storing the optimal state trajectory, 7 is a latch circuit (LATI) for storing each state at a target time,
Reference numeral 6 denotes a time-varying gain generator (ROMI) for generating an optimal manipulated variable for moving to the target state - at time T by specifying movement time T. Further, reference numeral 5 in the figure is a multiplier (MLYI) for generating a corrected manipulated variable from each state at the target time and the time-varying gain generated from the time-varying gain generator 6, and reference numeral 4 is a former optimal manipulated variable. and an adder (ADD) to generate a new optimal manipulated variable from the corrected manipulated variable.
I). 20 is A, 82 encoders 15
．． 16, waveform shaper 17, and UP/DovN counter 18
The servo motor 14 and each state (position, speed) of the controlled object driven by the motor 14 detected by a state detection system consisting of a state estimation circuit 19 and a system unit 1 constituted by a well-known microcomputer, etc.
A subtracter (SUBI) calculates the difference from the target state %e commanded by the subtractor (SUBI) to move the origin of each state, and reference numeral 12 is a subtracter (SUBI) that generates the deviation between each state after the origin is moved and the optimal state trajectory.
SUB2). In the figure, reference numeral 10 is a time-invariant gain generator (ROM2) that stores the feedback gain in a steady state, and reference numeral 11 is a multiplier (MLY2) that generates a correction operation amount from the state deviation and the time-invariant gain. 9
is an adder (ADD2) that adds a correction operation amount to the optimum operation amount series to generate an operation amount to be applied to the motor 14.

Incidentally, reference numeral 2 in the figure is an initial data generator (ROM 3) that stores initial data of the optimum operation amount obtained from the model in order to reduce the number of times of repeated learning.

Also, SW1 in the figure. SW2. SW3 is a changeover switch for switching between a learning system (A side) and a servo control system (B side) of the servo control device, and is operated by a learning/servo switching command signal from the system unit 1.

Here, the procedure for calculating the above-mentioned time-varying gain will be briefly explained.

Controllable and stable discrete time n-th order linear system (m-th order input)
1) from the initial state x0 to the target state Xs in time T = N-
The boundary conditions that transition at Ts (Ts: sampling time) are: X(0)=X, , X(N)=Xa=O(2)
At this time, the following quadratic form evaluation index (PI) is minimized by considering the amount of heat generated by the servo motor.

Here, A is an n×n state transition matrix, B is an n×m control input row example, R is an m×m input weight matrix, and X is an n-th state state vector.

εuler of the optimal control interval that minimizes this evaluation index
The -Lagrange equation is: and becomes a two-point boundary value problem. Note that P(K) is an n-th auxiliary variable vector.

Also, the manipulated variable U (K) is calculated from equation (4), U (K) = -
E-13-P(K)-(5).

Therefore, optimal control can be achieved by finding the auxiliary variable vector P (K) so as to satisfy the boundary conditions.

Here, if equation (4) is rewritten using the relationship between the initial state x0 and the target state Xe, it can be expressed as follows, where M, 0. M engineering 22M2□ are respectively, and the relationship of the auxiliary variable vector P (K) is uJ (K) = RB P (K) · Ke. Therefore, the 1 time-varying gain G(K) is G(K)=-M' [3
It is obtained from TP(K).

Note that the time-varying gain calculated from the above equation is stored in the time-varying gain generator (ROMI) 6 shown in FIG.

Now, as an example, the details of the control operation when the servo control device having the configuration shown in FIG. 1 is applied to positioning control will be explained.

First, when the system unit 1 is started, the initial data of the optimum operation amount stored in the initial data generator (ROM 3) 2 is copied to the function generator (RAM 1) 3, and then the learning/servo changeover switch SWI, SW2. SW3
is activated by a learning/servo switching command signal from system unit 1 and is connected to learning side A. Changeover switch SWI.

SW2. When SW3 is connected to learning side A, function generation W
The contents of (RAMI) 3 are sent to the servo amplifier 13, and the servo motor 14 is driven. When the servo motor 14 is driven, the mechanical system driven by the servo motor 14 is activated, and the encoder (A) installed as a 1-position sensor 15. Position information is detected by the U/D counter 18 based on pulses sent from the encoder (B) 16. Also, at the same time, speed information is detected by the state estimator 19.

On the other hand, the subtracter (SUBI) 20
The target state (position, velocity) commanded from the target state and the current state of the mechanical system (position, velocity) obtained via the state estimator 19.
, move the origin of the state, and modify the state to match the target state. As a result, all targets are unified into the O state. Each state after the origin is moved is stored in the function generator (RAM2) 8 as an optimal state trajectory, and the state error amount at the target time is latched by the latch circuit (LATI) 7.

On the other hand, the time-varying gain generator (ROMI) 6 selects a time-varying gain coefficient according to the movement time T commanded by the system unit 1, and transmits it to the multiplier (MLYI) 5. The multiplier (MLYI) 5 outputs the state error amount latched by the latch circuit (LATI) 7 and the time-varying gain generator (ROM).
I) Multiply by the time-varying gain coefficient from 6 to generate a corrected manipulated variable. Adder (ADDI) that receives this modified operation amount
4 corrects the old optimal manipulated variable stored in the function generator (RAMI) 3 by the corrected manipulated variable, generates a new optimal manipulated variable, and generates the optimal manipulated variable stored in the function generator (RAMI) 3. refresh.

The above is the learning system operation of the servo control device according to the present invention, and this learning system operation is repeatedly attempted until the state error amount at the target time reaches the vicinity of the origin.

As described above, in the servo control device according to the present invention, the optimum operation amount and state trajectory are determined by repeated trial operations of the learning system so that the preset model matches the actual state of the mechanical system at each target time. Fixed. Therefore, after learning, the error between the model and the actual state is reduced, and it is possible to always store the optimal operation amount and state trajectory in consideration of steady disturbances.

Next, after the trial learning is completed, the learning/servo switching command signal sent from the system unit 1 connects the learning/servo switching switches swi, sw2, and SW3 to the servo side, and the closed loop performs real-time state feedback control. A control system is configured.

Once this closed loop control system is configured, the encoder (A)
15 and the encoder (B) 16 and transmitted via the state estimation circuit 19, the current state of the mechanical system whose origin has been moved, and the function generator (RAM) generated by learning.
2) A zero multiplier (M
LY2) 11 uses the state deviation generated from the subtracter (SUB2) 12 and the time-invariant gain generator (ROM2) 10 in advance.
Multiply by the steady state feedback gain, which is memorized.

Generate a correction operation amount. And adder (ADD2)
9 adds this correction operation amount to the optimum operation amount series generated by learning and stored in the function generator (RAMI) 3 to generate an operation amount to be applied to the servo motor 14, and applies the operation amount to the servo motor through the servo amplifier 13. 14. Thereafter, real-time state feedback control of the servo motor 14 is performed based on the optimal state trajectory and optimal operation sequence generated by the learning system, and as a result, the trajectory can be quickly adjusted even in the face of sudden disturbances. Can be corrected.

The operation of the servo control device according to the present invention has been described above. By using the servo control device according to the structure of the present invention, the following effects can be obtained.

(1) As shown in the graph of FIG. 2, the specified target state Xs can be reached within the specified state transition time T as long as it is within the range of constraint conditions such as the amount of operation.

The graph in FIG. 2 shows the operation amount U and speed ω with respect to the time elapsed from the start of control to the specified state transition time T (time required to transition to the target state) before and after learning. , is a graph showing changes in position θ.

(2) As is clear from Figure 2, by repeatedly performing trials using the learning system in advance, the optimal manipulated variable sequence and optimal state trajectory that always minimizes energy can be generated in response to secular changes in the controlled object. Can be done.

(3) The servo control device according to the present invention has an additional feedforward operation, which reduces the burden on the closed loop control system, making it possible to apply stronger feedback compared to conventional control devices, resulting in disturbance suppression characteristics. The state of the controlled object accurately follows the output of the function generator.

(4) Controlled objects subject to gravity load or friction load, e.g.
In the case of manipulators, robot arms, etc., the nonlinearity of gravity and friction is too large for conventional linear servo systems, so
Although it was necessary to separately prepare compensation data in advance, this control device does not require compensation data because it can generate an optimal trajectory by itself through repeated trials using a learning system.

The present invention has been described above based on the illustrated embodiments, but the components of the servo control device according to the present invention are not limited to those shown in FIG. 1, and can be modified in various ways depending on the object to be controlled. .

(Effects of the Invention) As explained above, according to the servo control device of the present invention, the response characteristics during transient response of the servo control system can be improved through repeated learning, and overshoot occurring in linear feedback control can be reduced. , the settling time can be accelerated.

Therefore, according to the present invention, high-speed and high-precision positioning of a servo motor and a mechanical system driven by the servo motor, high-speed start-up during constant speed control and constant torque control,
High-speed stopping is possible.

Further, according to the servo control device according to the present invention, the servo motor can be driven with minimum energy during the specified settling time given from the design specifications.

[Brief explanation of the drawing]

1wi is a schematic configuration diagram of a servo control device showing an embodiment of the present invention, and FIG. 2 is a graph showing control states before and after learning by the servo control device having the configuration shown in FIG. It is a graph showing changes in the manipulated variable U, the speed ω1, and the position θ with respect to the elapsed time t from the start of control after learning until reaching a prespecified state transition time T. 1... System unit, 2... Initial data generator, 3.8... Function generator, 4, 9... Adder, 5.11... Multiplier, 6 ... Time-varying gain generator, 7... Latch circuit, 10... Time-invariant gain generator, 12゜20... Subtractor, 13... Servo amplifier, 14...・Servo motor, Is, 16・
... Encoder, 17... Waveform shaper, 18...
...U/D counter, 19...state estimation circuit, SW
l, SW2. SW3...Learning/servo changeover switch. 2nd session 1:1

Claims (1)

[Claims] In a digital servo control device that performs position control and speed control of a mechanical system driven by a motor, the optimal operation amount series and optimal state trajectory for the mechanical system driven by the motor are constantly determined in the final step without a real-time feedback loop. A learning system comprising a means for generating control by repeatedly trying control based on state deviation, and a means for generating an optimal manipulated variable series for a predetermined model in order to reduce the number of trials, and the learning system. A servo control device comprising: a closed-loop servo control system that performs real-time state feedback control of a servo motor based on an optimal state trajectory and an optimal manipulated variable series generated by the system.