METHOD FOR CONTROLLING POSITIONING OF ACTUATOR COMPRISING WAVE GEAR DEVICE

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to a method for controlling positioning of an actuator for reducing the rotational output of a motor via a wave gear device and providing output from a load shaft. More specifically, the present invention relates to a positioning control method of an actuator that is equipped with a wave gear device that makes it possible to use a strict linearization technique to inhibit any reduction in the accuracy with which positioning of the load shaft is controlled, as caused by the non-linear spring characteristics of the wave gear device.

2. Description of the Related Art

There is known in the art an actuator 1 in which a wave gear device 3 is employed as a drive for outputting a decelerated rotational output of a motor 2, as shown in FIG. 9. A controller 4 known in the art for controlling positioning of the actuator 1 of the above configuration performs a semi-closed loop control wherein the rotational position and rotational speed of a motor shaft 5 are detected by a sensor 6 attached to the motor shaft 5; and the rotation of the load shaft 7, which is the drive output shaft, is controlled based on the sensor readings. The characteristics of the wave gear device 3 greatly influence the positioning control characteristics of the load shaft 7 in such semi-closed loop control systems because driving of the motor 2 is not controlled by directly detecting rotational information relating to the load shaft 7.

Non-linear elastic deformation occurs between the input and output in the wave gear device when a load torque is applied, and is one factor preventing the load shaft from being controlled with a high degree of accuracy. The effects of the non-linear spring characteristic must be accounted for in order to achieve high-accuracy control of the load shaft.

Strict linearization of the input-output relationship is known as a control method for controlling non-linear elements. Strict linearization is a technique wherein a linearization feedback α(x) and an input conversion β(x) are performed as shown in FIG. 1, and linearization is accomplished by setting α(x) and β(x) so that the characteristics of the output y will be dny/dtn=v from the input v of the expansion system, which includes α(x) and β(x) (Non-Patent Reference 1).

[Prior Art Reference]
[Non-Patent Reference]

[Non-Patent Reference 1] “Non-Linear System Theory”, Ishijima et al., Society of Instrument and Control Engineers, CORONA PUBLISHING CO., LTD., p. 141-168, 1993.

SUMMARY OF THE INVENTION

An object of the present invention is to use, in a semi-closed loop system of an actuator for reducing the rotational output of a motor via a wave gear device and providing output from a load shaft, a strict linearization technique to make it possible to inhibit any reduction in the accuracy with which positioning of the load shaft is controlled, as caused by the non-linear spring characteristics of a wave gear device for which definite analysis and control techniques are not established.

To accomplish the above object, according to the present invention, there is provided a method for controlling positioning of an actuator comprising reducing a rotational output of a motor via a wave gear device and transmitting the output from a load shaft, and controlling positioning of the load shaft based on a rotational position and speed of a motor shaft of the motor, the method being characterized in comprising:

performing non-linear spring compensation for inhibiting a reduction in the accuracy with which positioning of the load shaft is controlled, as caused by non-linear elastic deformation in relation to a load torque of the wave gear device; and

in the non-linear spring compensation,

constructing a plant model that has been linearized using a strict linearization technique from the actuator to be controlled, and defining the linearization feedback α(x) and input conversion β(x) as well as the characteristics of the expansion system from input v to output y according to formulae (A), (B), and (C), respectively;

$\begin{matrix} [FORMULA A] \\ α (x) \cdot x = - \frac{J_{m} J_{l} N}{D_{g} K_{t}} {\begin{matrix} \begin{matrix} \begin{matrix} \frac{1}{J_{l}} \frac{\partial}{\partial t} τ_{g} (θ_{tw}) - \\ (\frac{D_{l} + D_{g}}{J_{l}^{2}} + \frac{D_{g}}{J_{m} J_{l} N^{2}}) τ_{g} (θ_{tw}) + \end{matrix} \\ (\frac{{(D_{l} + D_{g})}^{2}}{J_{l}^{2}} + \frac{D_{g}^{2}}{J_{m} J_{l} N^{2}}) ω_{l} - \end{matrix} \\ (\frac{D_{l} D_{g} + D_{g}^{2}}{J_{l}^{2} N} + \frac{D_{m} D_{g} N^{2} + D_{g}^{2}}{J_{m} J_{l} N^{3}}) ω_{m} \end{matrix}} & (A) \\ [FORMULA B] \\ β (x) \cdot υ = \frac{J_{m} J_{l} N}{D_{g} K_{t}} υ & (B) \\ [FORMULA C] \\ \frac{\partial^{3} y}{\partial t^{3}} = υ & (C) \end{matrix}$

measuring the non-linear elastic deformation in relation to the load torque of the wave gear device;

defining the non-linear spring model τg (θtw) using a cubic polynomial as shown in formula (D), with the constant defined as zero, to be able to recreate the measurement results; and

[FORMULA D]

τ_g(θ_tw)=K_g3θ_tw³+K_g2θ_tw²+K_g1θ_tw (D)

respectively inputting into a semi-closed loop control system for controlling the positioning of the load shaft, as a feed-forward current command and a feed-forward motor position command, the current input into the plant model and the motor position of the plant model when a load acceleration command is a command value.

Effects of the Invention

According to the present invention, a feed-forward non-linear spring compensation system based on a strict linearization technique in relation to the non-linear spring characteristics between the input and output shafts of a wave gear device of an actuator is constructed as a non-linear spring characteristic compensation method in the wave gear device. Over shooting of the load shaft is thereby reduced, and the load shaft can be smoothly and accurately stabilized at the target position.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block schematic diagram depicting a strict linearization technique;

FIG. 2 is a graph depicting a non-linear spring model;

FIG. 3 is a block schematic diagram depicting the feed-forward non-linear spring compensation system according to the present invention;

FIG. 4 is a graph depicting a command waveform employed in an experiment for confirming the non-linear spring compensation effect;

FIG. 5 is a graph depicting positioning response in an experiment for confirming the non-linear spring compensation effect;

FIG. 6 is a graph depicting positioning response in an experiment for confirming the non-linear spring compensation effect;

FIG. 7 is a graph depicting positioning response in an experiment for confirming the non-linear spring compensation effect;

FIG. 8 is a graph depicting positioning response in an experiment for confirming the non-linear spring compensation effect;

FIG. 9 is a schematic block diagram depicting an actuator comprising a wave gear device to be controlled by the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS
1. Strict Linearization Technique
(1.1 Description of the Formulae)

A strict linearization technique is one in which the object to be controlled is linearized by deriving a linearization feedback α(x) and an input conversion P(x) based on a physical equation comprising a non-linear function. The meanings of the symbols in each of the formulae below are as follows.

(Key)

θ_m: motor position

θ_l: load position

ω_m: motor speed

ω_l: load speed

x: state quantity

x=[θ_mθ_lω_mω_l]

α(x): linearization feedback

β(x): input conversion

u: input (of the state formula)

i: electrical current

$θ_{tw} = \frac{θ_{m}}{N} - θ_{l}$

τ_g(θ_tw): nonlinear spring model

N: reduction ratio

K_t: torque constant

J_m: motor inertia

J_l: load inertia

D_m: motor viscosity coefficient of friction

D_l: load viscosity coefficient of friction

K_g: spring constant

D_g: spring viscosity coefficient of friction

K_g3, K_g2, K_g1: non-linear spring model coefficient

v: input of expansion system

y: output of expansion system

i*_ref: FF current command

θ*_m: FF motor position command

θ*_t: increase acceleration command

First, the characteristics of the object to be controlled are expressed using a non-linear differential equation. The global characteristics of an actuator 1 to be controlled, depicted in FIG. 9, are expressed in a two-inertia model where twisting occurs between the input and output shafts of a wave gear device. The state equation of the two-inertia model expressing the wave gear device becomes formula (1) when the input u is set equal to the current i and the state quantity x is set equal to load position θ_l, load speed ω_l, motor position θ_m, and motor speed ω_m. Formula (2) is obtained when the right side of formula (1) is expanded.

$\begin{matrix} [FORMULA 1] \\ \frac{\partial}{\partial t} [\begin{matrix} θ_{l} \\ ω_{l} \\ θ_{m} \\ ω_{m} \end{matrix}] = [\begin{matrix} 0 & 1 & 0 & 0 \\ - \frac{K_{g}}{J_{l}} & - \frac{D_{l} + D_{g}}{J_{l}} & \frac{K_{g}}{J_{l} N} & \frac{D_{g}}{J_{l} N} \\ 0 & 0 & 0 & 1 \\ \frac{K_{g}}{J_{m} N} & \frac{D_{g}}{J_{m} N} & - \frac{K_{g}}{J_{m} N^{2}} & - \frac{D_{m} N^{2} + D_{g}}{J_{m} N^{2}} \end{matrix}] [\begin{matrix} θ_{l} \\ ω_{l} \\ θ_{m} \\ ω_{m} \end{matrix}] + [\begin{matrix} 0 \\ 0 \\ 0 \\ \frac{K_{t}}{J_{m}} \end{matrix}] i & (1) \\ [FORMULA 2] \\ \frac{\partial}{\partial t} [\begin{matrix} θ_{l} \\ ω_{l} \\ θ_{m} \\ ω_{m} \end{matrix}] = [\begin{matrix} ω_{l} \\ \frac{K_{g}}{J_{l}} (\frac{θ_{m}}{N} - θ_{l}) - \frac{D_{l} + D_{g}}{J_{l}} ω_{l} + \frac{D_{g}}{J_{l} N} ω_{m} \\ ω_{m} \\ - \frac{K_{g}}{J_{m} N} (\frac{θ_{m}}{N} - θ_{l}) + \frac{D_{g}}{J_{m} N} ω_{l} - \frac{D_{m} N^{2} + D_{g}}{J_{m} N^{2}} ω_{m} \end{matrix}] + [\begin{matrix} 0 \\ 0 \\ 0 \\ \frac{K_{t}}{J_{m}} i \end{matrix}] & (2) \end{matrix}$

According to the present invention, the linear spring model Kg(θm/N−θl) is expressed in a first-order non-linear differential equation, shown in formula (3), by substitution with a non-linear spring model τg(θtw) wherein spring torque varies according to the twisting θtw=θm/N−θl.

$\begin{matrix} [FORMULA 3] \\ {\begin{matrix} \frac{\partial θ_{l}}{\partial t} = ω_{l} \\ \frac{\partial ω_{l}}{\partial t} = \frac{1}{J_{l}} τ_{g} (θ_{tw}) - \frac{D_{l} + D_{g}}{J_{l}} ω_{l} + \frac{D_{g}}{J_{l} N} ω_{m} \\ \frac{\partial θ_{m}}{\partial t} = ω_{m} \\ \frac{\partial ω_{m}}{\partial t} = - \frac{1}{J_{m} N} τ_{g} (θ_{tw}) + \frac{D_{g}}{J_{m} N} ω_{l} - \frac{D_{m} N^{2} + D_{g}}{J_{m} N^{2}} ω_{m} + \frac{K_{t}}{J_{m}} i \end{matrix} & (3) \end{matrix}$

The load position θl, which is the final control amount, is then expanded from formula (3) while derived one order at a time. The first derivative dθl/dt, second derivative d²θl/dt², and third derivative d³θl/dt³of θl become formulas (4), (5), and (6), respectively; and the current i, which is the control input, appears in the formula at the third derivative.

$\begin{matrix} [FORMULA 4] \\ \frac{\partial 0_{l}}{\partial t} = ω_{l} & (4) \\ [FORMULA 5] \\ \begin{matrix} \frac{\partial^{2} θ_{l}}{\partial t^{2}} = \frac{\partial}{\partial t} \frac{\partial θ_{l}}{\partial t} \\ = \frac{\partial ω_{l}}{\partial t} \\ = \frac{1}{J_{l}} τ_{g} (θ_{tw}) - \frac{D_{l} + D_{g}}{J_{l}} ω_{l} + \frac{D_{g}}{J_{l} N} ω_{m} \end{matrix} & (5) \\ [FORMULA 6] \\ \begin{matrix} \frac{\partial^{3} θ_{l}}{\partial t^{3}} = \frac{\partial}{\partial t} \frac{\partial^{2} θ_{l}}{\partial t^{2}} \\ = \frac{\partial}{\partial t} \frac{1}{J_{l}} τ_{g} (θ_{tw}) - \frac{\partial}{\partial t} \frac{D_{l} + D_{g}}{J_{l}} ω_{l} + \frac{\partial}{\partial t} \frac{D_{g}}{J_{l} N} ω_{m} \\ = \frac{1}{J_{l}} \frac{\partial}{\partial t} τ_{g} (θ_{tw}) - \frac{D_{l} + D_{g}}{J_{l}} \frac{\partial ω_{l}}{\partial t} + \frac{D_{g}}{J_{l} N} \frac{\partial ω_{m}}{\partial t} \\ = \frac{1}{J_{l}} \frac{\partial}{\partial t} τ_{g} (θ_{tw}) - \frac{D_{l} + D_{g}}{J_{l}} \\ (\frac{1}{J_{l}} τ_{g} (θ_{tw}) - \frac{D_{l} + D_{g}}{J_{l}} ω_{l} + \frac{D_{g}}{J_{l} N} ω_{m}) + \\ \frac{D_{g}}{J_{l} N} (- \frac{1}{J_{m} N} τ_{g} (θ_{tw}) + \frac{D_{g}}{J_{m} N} ω_{l} - \frac{D_{m} N^{2} D_{g}}{J_{m} N^{2}} ω_{m} + \frac{K_{t}}{J_{m}} i) \\ = \frac{1}{J_{l}} \frac{\partial}{\partial t} τ_{g} (θ_{tw}) - (\frac{D_{l} + D_{g}}{J_{l}^{2}} + \frac{D_{g}}{J_{m} J_{l} N^{2}}) τ_{g} (θ_{tw}) + \\ (\frac{{(D_{l} + D_{g})}^{2}}{J_{l}^{2}} + \frac{D_{g}^{2}}{J_{m} J_{l} N^{2}}) ω_{l} - \\ (\frac{D_{l} D_{g} + D_{g}^{2}}{J_{l}^{2} N} + \frac{D_{m} D_{g} N^{2} + D_{g}^{2}}{J_{m} J_{l} N^{3}}) ω_{m} + \frac{D_{g} K_{t}}{J_{m} J_{l} N} i \end{matrix} & (6) \end{matrix}$

While the load position θl in relation to any input i can be obtained by solving formula (6), formula (6) is a third-order non-linear differential equation, and obtaining a generalized solution is difficult. The electrical current i is calculated from formula (6) as formula (7) using input v and state quantities θl, ωl, θm, and ωm of the expanded system. From formula (7), the linearization feedback α(x) and input conversion β(x) become as shown in formulae (8) and (9) respectively. In this case, the characteristics of the expansion system from the input v to output y become as shown in formula (10).

$\begin{matrix} [FORMULA 7] \\ i = - \frac{J_{m} J_{l} N}{D_{g} K_{t}} {\begin{matrix} \begin{matrix} \begin{matrix} \frac{1}{J_{l}} \frac{\partial}{\partial t} τ_{g} (θ_{tw}) - \\ (\frac{D_{l} + D_{g}}{J_{l}^{2}} + \frac{D_{g}}{J_{m} J_{l} N^{2}}) τ_{g} (θ_{tw}) + \end{matrix} \\ (\frac{{(D_{l} + D_{g})}^{2}}{J_{l}^{2}} + \frac{D_{g}^{2}}{J_{m} J_{l} N^{2}}) ω_{l} - \end{matrix} \\ (\frac{D_{l} D_{g} + D_{g}^{2}}{J_{l}^{2} N} + \frac{D_{m} D_{g} N^{2} + D_{g}^{2}}{J_{m} J_{l} N^{3}}) ω_{m} \end{matrix}} + \frac{J_{m} J_{l} N}{D_{g} K_{l}} υ & (7) \\ [FORMULA 8] \\ α (x) \cdot x = - \frac{J_{m} J_{l} N}{D_{g} K_{t}} {\begin{matrix} \begin{matrix} \begin{matrix} \frac{1}{J_{l}} \frac{\partial}{\partial t} τ_{g} (θ_{tw}) - \\ (\frac{D_{l} + D_{g}}{J_{l}^{2}} + \frac{D_{g}}{J_{m} J_{l} N^{2}}) τ_{g} (θ_{tw}) + \end{matrix} \\ (\frac{{(D_{l} + D_{g})}^{2}}{J_{l}^{2}} + \frac{D_{g}^{2}}{J_{m} J_{l} N^{2}}) ω_{l} - \end{matrix} \\ (\frac{D_{l} D_{g} + D_{g}^{2}}{J_{l}^{2} N} + \frac{D_{m} D_{g} N^{2} + D_{g}^{2}}{J_{m} J_{l} N^{3}}) ω_{m} \end{matrix}} & (8) \\ [FORMULA 9] \\ β (x) \cdot υ = \frac{J_{m} J_{l} N}{D_{g} K_{t}} υ & (9) \\ [FORMULA 10] \\ \frac{\partial^{3} y}{\partial t^{3}} = υ & (10) \end{matrix}$

(1.2 Non-Linear Spring Model)

It follows from Eq. (8) that the non-linear spring model τg(θtw) and its first-order derivative dτg(θtw)/dt are needed for non-linear spring compensation; therefore, τg(θtw) must have a first-order derivative. Accordingly, in the present invention, the spring characteristics measured for the device as shown by line I in FIG. 2 are expressed by a third-order polynomial shown in formula (11), in which the constant term is set to zero, as a non-linear spring model τg(θtw); and these characteristics are shown by line II in FIG. 2. It follows from FIG. 2 that the spring characteristics measured for the device can be more accurately reproduced than with the linear spring shown by line III, despite hysteresis characteristics not being accounted for in the constructed non-linear spring model.

[FORMULA 11]

τg(θ_tw)=K_g3θ_tw³+K_g2θ_tw²+K_g1θ_tw (11)

2. FF Non-Linear Spring Compensation

The linearization feedback α(x) and the input conversion β(x) are derived from formulae (8) and (9), as described above. As can be seen in formula (8), however, the load position θl and load speed ωl are required for the linearization feedback α(x). The control system of the actuator to be controlled according to the present invention is a semi-closed loop control system, and data on the load shaft cannot be obtained.

As shown in FIG. 3, the entire block-line diagram shown in FIG. 1 is installed in the controller, and the current input i*_refand the motor position θ*m of the plant model are inputted respectively as the FF current command and the FF motor position command when the load acceleration command is set as the command value. As long as the characteristics of the plant model and the characteristics of the device are similar, the device will thereby operate similarly to the linearized plant model; and the non-linear spring characteristics can be compensated for.

3. Experiment for Confirming Compensation Effect

The feed-forward non-linear spring compensation method described above was incorporated in a controller of an actuator in the device, and the associated compensation effect was confirmed in a positioning experiment. The conditions of the experiment are shown in Table 1. In the 2-degrees-of-freedom control system based on the coprime factorization description according to the conventional method, the linear spring model (line III) of FIG. 2 was used, and a command in the form of the trapezoidal acceleration shown in FIG. 3 and transmitted by a lowpass filter was used in both the conventional method and the method according to the present invention.

TABLE 1

Conditions for the experiment for confirming

FF non-linear spring compensation effect

Conventional method
Proposed method

Control system
2-degrees-of-freedom
FF non-linear spring

control system
compensation + P-PI

control system

Load inertia ratio
2.95

Target position
36, 43.2, 43.56

[Load deg]

Command
Motor position
Load acceleration

The positioning response obtained for twelve consecutive runs of the load shaft at 36 load deg are shown in FIGS. 5 through 8. In each of the drawings, (a) shows the response results of the conventional method, and (b) shows the response results of the method according to the present invention. The horizontal broken lines are a load shaft stabilization range of ±30 load arc sec, a motor stabilization range of ±10 motor pulses, and a maximum current rating of ±0.64 A. As can be seen from the load position and motor position response in FIGS. 5 and 6, while there is an overshoot of approximately 0.05 load deg and 0.025 load deg in both the load shaft and the motor shaft, the overshoot is reduced in the method according to the present invention to approximately 0.02 load deg in the load shaft and 0.01 load deg in the motor shaft.

CONCLUSION

As described above, a feed-forward non-linear spring compensation system based on a strict linearization technique was constructed as a non-linear characteristic compensation method in a wave gear device in regard to a non-linear spring characteristic between the input and output shafts of the wave gear device, and the compensation effects were confirmed by experimentation on an actual device. The results confirmed that the overshoot of the load shaft can be reduced and smooth stabilization achieved by performing the non-linear spring compensation based on the strict linearization technique.

METHOD FOR CONTROLLING POSITIONING OF ACTUATOR COMPRISING WAVE GEAR DEVICE

Information

Publication Number

Date Filed

Date Published

Inventors

Original Assignees

CPC

US Classifications

International Classifications

Abstract

Description

Claims

Priority Claims (1)