JPH04324507A - Acceleration/deceleration control method of drive device and robot system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention provides a motor torque-speed characteristic that realizes a desired operation in the shortest possible time without causing insufficient motor torque in a multi-axis mechanism such as a robot driven by a motor. and an optimal motor acceleration/deceleration control method that takes into account mechanical characteristics.

0002

2. Description of the Related Art Conventionally, a method for controlling the acceleration and deceleration of a drive device disposed at each joint of a robot, which is a typical example of a multi-axis mechanism, is described in 1) Japanese Patent Application Laid-Open No. 62-80706. A method was described in which the axial acceleration is determined based on the required torque multiplier coefficient in the direction and a speed pattern is generated.

2) In Japanese Patent Application Laid-open No. 114317/1983, the required torque of the motor determined from the dynamic equation with a predetermined speed pattern exceeds the torque that can be generated by the motor at the operating shaft speed. In such cases, methods to reduce acceleration were described.

3) In Japanese Patent Application Laid-open No. 62-3311, based on the acceleration/deceleration parameters that are tabulated for the robot's motion start position and motion distance, For operating conditions, a method was described in which acceleration/deceleration parameters are determined by interpolation and speed pattern planning is performed.

[0005] Furthermore, in 4) Japanese Patent Laid-Open No. 60-201408 and No. 61-271501, in order to reduce residual vibration when the robot stops operating, a cycloidal velocity pattern is used at the time of stopping and starting. The method used was described.

[0006] Also, 5) As described in Japanese Patent Application Laid-Open No. 2-106284, there is a method of minimizing the operation time based on the trapezoidal speed pattern and taking into account the speed limit and acceleration limit of each joint of the robot. was stated.

[0007] Also, 6) As described in Machine Design Vol. 33, No. 3, 1989, pages 64 to 92, flexibility is provided in selecting the motion curve of the follower moved by the cam. Therefore, a universal cam curve was described in which the time width of seven sections each consisting of a sine curve, a cosine curve, and a straight line can be arbitrarily set.

[0008]

[Problems to be Solved by the Invention] The first prior art ignores nonlinear terms (centrifugal and Coriolis components) in the dynamic equation that expresses motor torque in a multi-axis mechanism. There is a problem in that it is not possible to optimize the acceleration/deceleration pattern during high-speed operation of a multi-axis mechanism having a mass distribution in which the influence of

[0009] In the second prior art, the speed pattern is kept constant and the maximum speed and acceleration/deceleration time are varied to create an optimal pattern.
Since the engine is trying to find the desired speed, if the speed pattern used is inappropriate, the acceleration/deceleration time will be long.

Furthermore, since the third conventional technique deals with a small number of operating condition parameters and acceleration/deceleration parameters, it is difficult to obtain an operation that can minimize the operating time of the moving object. There's a problem.

Furthermore, in the fourth prior art, since the acceleration at the center speed during acceleration/deceleration is extremely large, there is a possibility that the motor may run out of torque during high acceleration/deceleration, making it difficult to determine the optimum acceleration/deceleration pattern. The problem is that you can't get it.

Furthermore, since the fifth prior art uses a trapezoidal velocity pattern, the acceleration at the time of stopping is not zero, so there is a problem that the amplitude of the residual vibration is large and lasts for a long time.

[0013] Furthermore, in the sixth prior art, since the speed pattern is set using the acceleration/deceleration time, maximum speed, and time width of each section as variable parameters, the setting of the time width of each of the seven sections is The problem is that there are many variables and it is extremely difficult.

An object of the present invention is to provide a speed pattern generation method that can minimize the operating time during high acceleration/deceleration operations of a drive device and a multi-axis mechanism driven by the drive device.

[0015]

[Means for Solving the Problems] In order to achieve the above objects, the present invention (I) uses an acceleration/deceleration curve that allows flexibility in speed pattern generation, (II) minimizes operating time, Generate a speed pattern that simultaneously minimizes the torque margin of the motor and minimizes residual vibration when the motor stops. (III) Using a simple method to determine operating condition parameters and optimal acceleration/deceleration parameters. This method takes three measures: linking and easily calculating the optimum acceleration/deceleration parameters for any operating conditions. Each will be specifically explained below.

(I) In order to obtain an acceleration/deceleration curve that allows flexibility in speed pattern generation, an exponential sine wave curve shown in Equation 2 was used.

[0017]

[Math 2]

Here, θ: shaft speed, θm: maximum speed, m
: index, th: acceleration/deceleration period, t: time, θm,
The three parameters m and th are acceleration/deceleration parameters. This corresponds to the case where m = 2 for the conventionally known cycloid curve and the case where m = 1 for the single sinusoidal curve, and the existing sine wave system acceleration/deceleration curve has only acceleration/deceleration parameters θm and th. Compared to the previous version, m was added to increase flexibility.

(II) When generating the speed pattern, the above-mentioned exponential sine wave acceleration/deceleration curve was used to determine the shaft speeds of the acceleration section, constant velocity section, and deceleration section as shown in Equation 1.

Here, when the operating distance of the moving body is Δθ, the acceleration distance αΔθ and the deceleration distance βΔθ are shown as in equation 3, and among the acceleration and deceleration parameters, the acceleration period th1 and the deceleration period th2 are Since the acceleration distance α and specific deceleration distance β are determined uniquely, m1, m2 are used as acceleration/deceleration parameters.
, θm, α, and β were used.

[0021]

[Math 3]

[0022] Here, Γ represents a gamma function.

Furthermore, when driving a load at high speed with a motor, it is desirable to minimize the operating time as long as the motor does not run out of torque. The optimum acceleration/deceleration parameters were determined and used for the conditions (load inertia moment, operating distance, etc.).

(III) In associating operating condition parameters with optimal acceleration/deceleration parameters, in order to relate multiple variables, (i) table data utilization method, (ii) template - Bull data interpolation method, (iii) least squares polynomial approximation method, and (iv) neural network learning method.

[0025]

[Operation] High-speed moving objects such as robot arms use their drive motors.
By applying electrical energy to the motor, it is converted into mechanical energy and driven. When a speed command is given to the drive motor control device, a motor torque command is generated based on the deviation from the speed detection amount obtained from the drive motor speed detector, is amplified, and energizes the motor. be done.

The operation required of a high-speed moving object is to position it as quickly as possible from the operation start position to the operation end position as shown in FIG. 26, and to reduce the residual vibration amplitude and its damping as much as possible as shown in FIG. The goal is to save time.

Hereinafter, means for realizing such an operation will be explained using FIG. 1 and FIGS. 28 to 38. FIG. 1 shows a speed pattern using an exponential sine wave acceleration/deceleration curve, FIG. 28 shows an exponential sine wave curve, and FIG.
30 shows the relationship between the maximum value of the dimensionless parameter of the exponential sine wave acceleration/deceleration curve and the index, and FIG. 31 shows the relationship between the dimensionless parameter of the exponential sine wave acceleration/deceleration curve and the index. The relationship between the dimensionless time that takes the maximum acceleration of the acceleration/deceleration curve and the index is shown, FIG. 32 shows an external view of the multi-axis mechanism, and FIG.
Fig. 34 shows the generated torque-speed characteristics of the motor, and Fig. 34 shows the time change of the required torque under optimum driving conditions. Fig. 35
shows the configuration of the neural network, and FIG. 36 shows the configuration of the neural network.
It shows the convergence characteristics of acceleration/deceleration parameter errors.

First, the characteristics of the exponential sine wave curve will be explained using FIG. 1 and FIGS. 28 to 31. The speed pattern shown in equation 1 is shown in FIG. In order to investigate the characteristics of the exponential sine wave curve (see Equation 2, Figure 28) used in the acceleration and deceleration sections in this pattern as an acceleration/deceleration curve, we investigated the Dimensional velocity V, dimensionless acceleration A, and dimensionless jump J are determined as shown in Equation 4.

[0029]

[Math 4]

Here, Vm=√πΓ((m+2)/2)
/Γ((m+1)/2), dimensionless time τ=2t/th
It is. Since this dimensionless parameter indicates a dimensionless quantity with an average value, the smaller the maximum value is, the smaller the impact is with less change.

FIG. 29 shows the time-varying characteristics of V, A, and J. This property has a large index dependence. V, A, J
The relationship between the maximum values Vm, Am, Jm and the index is shown in FIG. Also, the dimensionless time τA (Equation 5
) is shown in FIG. 31, indicating that the closer τA is to 0.5, the more maximum driving force is required at high speed.

[0032]

[Math 5]

From the above characteristics, the following relationship exists between the exponent m of an exponential sine wave acceleration/deceleration curve and its suitability as an acceleration/deceleration curve.

(i) m<1: The acceleration becomes infinite at the start of operation (τ=0) and at the end of operation (τ=1), and infinite motor driving force is required, so acceleration and deceleration are It is not preferable as a curve.

(ii) m=1: At the start of operation (τ=0, speed zero), the acceleration takes a finite maximum value and the maximum driving force is required of the motor, so normally the motor Considering that the maximum driving force can be generated at zero, this is suitable as an acceleration curve, and since Vm is small, the acceleration time can be shortened.

(iii) 1<m<2: The acceleration A at the start and end of the motion is zero, and the jump J is infinite. This curve is suitable as an acceleration curve because the speed at which the acceleration is maximum is in the low speed range, and it is also suitable as a deceleration curve because motors are usually less likely to generate insufficient driving force in the low speed range.

(iv) m≧2: Since the acceleration A and the jump J at the start and end of the operation are zero, this is suitable as a deceleration curve. However, the maximum acceleration time τA is 0.
Since the value is close to 5 and the maximum driving force is required at high speeds, the deceleration period th2 must be long in order to prevent the motor from running out of torque during high speed operation, resulting in a longer operating time. There is.

From the above, the acceleration index is m1≧1,
It is desirable to select m2>1 as the deceleration index.

Next, a method for determining the speed pattern of a high-speed moving body will be described with reference to FIGS. 32 to 34.

FIG. 32 is an external view of the n-axis robot.
A rotary drive motor is attached to each joint (each axis). The position of the tip is determined by combining the displacements θ of the drive motors of each joint. Here, j based on the rigid body model of the n-axis robot
The equation of motion of the shaft is generally expressed by Equation 6.

[0041]

[Math 6]

Here, rj: reduction gear reduction ratio, Tj: motor
torque, θj: motor shaft speed, θj: motor shaft acceleration, Hji: moment of inertia, Bjkl: centrifugal/Coriolis coefficient, Cj: gravitational moment, and Hji, Bjk
l and Cj are both functions of the axis position θj. When Expression 6 is written as a constant converted to the output shaft of the reducer (for example, Tj=rjTj, . . . ), it is expressed as Expression 7.

[0043]

[Math 7]

Here, the generated torque-speed characteristic when the maximum current is applied to the motor is shown as shown in FIG. 33 in terms of the output shaft of the reducer, and is described as shown in Equation 8.

[0045]

[Math. 8]

Further, the speed pattern of the j-axis is described as shown in Equation 9.

[0047]

[Math. 9]

[0048] Here, the n-axis robot is at point A (=

[Xa,
Ya, Za】) to point B (=

Find a speed pattern for each axis that will move to [Xb, Yb, Zb] in the shortest time and will not cause significant vibration after operation.

(Step 1) Find the axis-resolved coordinates (inverse kinematics solution) of points A and B.

(Step 2) Formulate the operation time of each axis.

When the specific acceleration distance αj and specific deceleration distance βj are set for each axis operating distance Δθjab, Equation 10 holds true, so the total operating time tj3 is expressed by Equation 11.

[0052]

[Math. 10]

[0053]

[Math. 11]

(Step 3) Determination of acceleration/deceleration parameters that minimize operating time to avoid torque shortage.

The sum f (see Equation 12) of the operating times of all axes is assumed to be the minimization evaluation function.

[0056]

[Math. 12]

[0057] Here, the following constraint conditions are set.

■Conditions that do not cause torque shortage (see FIG. 34 and Equation 13)

[0059]

[Math. 13]

■Acceleration/deceleration distance conditions 1-αj-βj≧0 0<αj<1, 0<βj<1 ■Guidelines for index selection mj1≧1, mj2≧1.2 (selected from mj2>1) ) ■ Constraints on axis maximum speed 0≦θjm≦θjm0 By solving this problem in such a way as to minimize the evaluation function considering the constraints, the optimal acceleration/deceleration parameters (mj1,
mj2, θjm0, αj, βj) are obtained.

Next, the operating condition parameters [Hji, Bjkl, Cj, rj,
θj (t=0), Δθj, Pj (θj)] and optimal acceleration/deceleration parameters [mj1, mj2, θjm0, αj, βj]
Figure 35 shows the means used to establish a simple relationship between
~Explained using FIG. 38. In the following description, for the sake of simplicity, they will be referred to as operating condition parameters (x1 to xs) and optimal acceleration/deceleration parameters (y1 to ym).

(I) Table data usage method In establishing the relationship between the above-mentioned optimal acceleration/deceleration parameters (y1 to ym) and operating condition parameters (x1 to xs), table data in the drive device control device is used. - Prepare pair data (x, y) in a table, and when a certain operating condition (xq) is given, the table data (x
Find the closest one among t). Since the y data is entered in the table, almost optimal acceleration/deceleration parameters can be easily determined. By preparing table data in detail, it becomes possible to obtain highly accurate interpolated values.

(II) Table data interpolation method The above optimum acceleration/deceleration parameters (y1 to ym) and operating condition parameters
When relating the data (x1 to xs), the relationship shown in Equation 14 is used.

[0064]

[Math. 14]

Here, pair data (x, y, dy/dx) is prepared in advance in a table in the drive control device, and when a certain operating condition (xq) is given, the table data is - Find the closest one among the data (xt). Since y, dy/dx data are entered in the table, by substituting these values into Equation 14, the optimum acceleration/deceleration parameter y can be determined. table day
By preparing the data in detail, it is possible to obtain highly accurate interpolated values.

(III) Least Squares Polynomial Approximation Method The optimum acceleration/deceleration parameters (y1 to ym) and the operating condition parameters (x1 to xs) are related by the polynomial shown in Equation 15.

[0067]

[Math. 15]

The polynomial coefficient ai is selected so that the least square error Δ2 shown in Equation 16 between the optimum value y and the r points is minimized.

[0069]

[Math. 16]

After the polynomial coefficients are determined, the optimum acceleration/deceleration parameters for any operating condition parameters can be easily calculated based on Equation 15.

Further, by dividing the operating condition parameters into a plurality of regions and deriving an approximate polynomial for each region, it is possible to calculate acceleration/deceleration parameters with higher precision.

(IV) Neural network learning method The above optimal acceleration/deceleration parameters (y1 to ym) and operating condition parameters (x1 to
Consider making connections using a physical network. The hierarchical neural network has a three-layer structure: an input layer 4, a hidden layer 5, and an output layer 6. In a learning step 8, relationships between each layer are learned, and in a reproduction step 9.
In this step, optimum acceleration/deceleration parameters for arbitrary operating condition parameters are determined based on the relationship obtained in the learning step. Calculations from the input layer 4 to the hidden layer 5 and from the hidden layer 5 to the output layer 6 are performed using the weighting coefficient shown in Equation 17 based on the deviation between the teacher data 10 and the output 11 determined in advance in steps 1 to 3 above. The configuration is such that the calculation shown in Equation 18 is performed on the weighting coefficient determined based on the modified algorithm.

[0073]

[Math. 17]

[0074]

[Math. 18]

Here, the function f is a sigmoid function shown by equation 19.

[0076]

[Math. 19]

FIG. 36 shows the state of convergence of output parameter errors due to modification of weighting coefficients in learning based on the above algorithm.

From now on, by using a neural network, relationships between multivariable input and output parameters can be established using a simple method. Furthermore, by dividing the operating condition parameters into a plurality of regions and performing the above learning for each region, more accurate approximation can be achieved.

[0079] Since both of the above methods (I) to (IV) involve errors in the relationship between input and output, the optimum acceleration/deceleration parameters are determined in the above-mentioned step 3, taking into consideration the margin for the error. By asking, under all operating conditions,
It is possible to easily set the optimum acceleration/deceleration parameters that will prevent the motor from running out of torque.

Furthermore, before executing the above-mentioned process, by performing a standard operation of the motor, measuring its required torque, and providing a step to identify the load inertia moment, more accurate optimal acceleration/deceleration parameters can be determined. - It becomes possible to determine the data.

[0081]

Embodiment A first embodiment of the present invention will be described with reference to FIGS. 1 to 9. This embodiment describes the relationship between operating condition parameters and optimal acceleration/deceleration parameters for a single-axis mechanism. Fig. 1 shows the speed pattern of the present invention, Fig. 2 shows the optimization procedure of the speed pattern, Fig. 3 shows the appearance of the single-axis mechanism and its control device, and Fig. 4 shows the operating time. 5 shows the convergence characteristics of the acceleration/deceleration index, FIG. 6 shows the convergence characteristics of the maximum speed, FIG. 7 shows the convergence characteristics of the specific acceleration/deceleration distance, and FIG. 8 shows the convergence characteristics of the acceleration/deceleration index. A comparison of the operating time with a known acceleration/deceleration curve is shown, and FIG. 9 shows a flowchart for determining the optimum acceleration/deceleration parameters for given operating condition parameters.

Here, as a single-axis mechanism to be treated, the inertial load body 1 is a direct drive motor, as shown in FIG.
This section deals with a mechanism that is rotationally driven by a motor (D/D motor) 2. The motor control device 3 has a function of generating and applying electrical energy to the motor to realize an operation according to the position command or speed command given from the outside. The motor speed pattern described in the present invention refers to the data that is given to the motor control device 3 as a speed command or the integral amount thereof is given as a position command.
This is an important technology that governs the acceleration/deceleration control of the inertial load body 1. Here, the reduction ratio r of the drive system is 1. Equation of motion of this mechanism, motor characteristics (example of hybrid D/D motor)
, and the acceleration/deceleration pattern (see Figure 1) are number 20, number 21,
and is shown by number 22.

[0083]

[Math. 20]

[0084]

[Math. 21]

[0085]

[Math. 22]

Here, the inertia moment of inertia H (takes a constant value regardless of the operating position θ) and the operating distance Δθ are given, and the acceleration/deceleration parameters (acceleration index m1, deceleration index m2
, maximum speed θm, specific acceleration distance α, specific deceleration distance β). Here, (m1, m2, θm, α, β) = (x
1, x2, x3, x4, x5), and the operation time (minimization evaluation function) f shown in Equation 23 is minimized under the constraint expression shown in Equation 24.

[0087]

[Math. 23]

[0088]

[Math. 24]

The first and second equations of the constraint equations shown in Equation 24 represent the balance between variables x6 and x7 corresponding to motor torque shortage and motor torque margin in the acceleration section and deceleration section, and the third equation are specific constant velocity distance x8, specific acceleration distance x4, specific deceleration distance x5
Equations 4 and 5 represent equality constraint expressions for determining phases x9 and x10 that minimize the torque margin in the acceleration section and the deceleration section. The possible ranges of each variable are as follows.

x1≧1, x2≧1.2, 0<x3≦θm0,0<
x4<1,0<x5<1 x6≧0,x7≧0,0<
x8<1, 0≦x9≦π/2, π/2≦x10≦π Here, when the operating distance Δθ and the moment of inertia of the inertial body H are given, the procedure for determining the optimal acceleration/deceleration parameters is as follows. It is shown in Figure 2. Details of each procedure shown in FIG. 2 are shown below. (I) Calculation of initial parameter values Determine provisional initial values x10 to x50 of acceleration/deceleration parameters,
x9 that satisfies the fourth and fifth constraint expressions in Equation 24
Find 0,x100. Next, the parameters shown in Equation 25 are modified so that the torque margins x6 and x7 of the accelerating section and the decelerating section are set to zero, and initial values are determined by repeating calculations until convergence.

[0091]

[Math. 25]

Note that the parameters are updated so that the variables x1 to x10 do not exceed the possible ranges mentioned above.

(II) Parameters are updated based on the evaluation function minimization parameter update number 26.

[0094]

[Math. 26]

(III) Motor torque margin minimization parameter correction procedure The acceleration/deceleration parameters obtained in step (II) do not satisfy the constraint equation shown in Equation 24. Therefore, the procedure (I
) Modify the parameters using the same procedure as described above. (I
V) Parameter convergence procedures (II) and (III) are repeated until the resulting acceleration/deceleration parameter change amount |Δxj|<ε (where ε: minute amount) and the parameters are converged.

Next, an example of calculation using the above optimization method will be shown. Motor characteristics (η=60.
63 (Nm√s/rad), γ=10.88 (rad/
s)) Figure 4 shows the convergence characteristics of the operating time when the load inertia moment H = 5 (kgm2) and the operating distance Δθ = 2π (rad). From now on, both the above-mentioned procedure (I) and procedure (IV) converge after about three repetitions. Further, the convergence characteristics of acceleration/deceleration index, maximum speed, and specific acceleration/deceleration distance are shown in FIGS. 5, 6, and 7, respectively. Furthermore, from a study with different initial parameters, the optimum acceleration/deceleration parameters are given as follows.

(1) Acceleration/deceleration index: m1=1.0 and m2=1.2, which correspond to the intersection of the prohibited boundary lines of m1 and m2, are best.

(2) Maximum speed: Set high enough to avoid torque shortage.

(3) Specific acceleration/deceleration distance: (α, β)=(0.
5,0.5) is uniquely determined by determining other parameters.

FIG. 8 and Table 1 show that the speed pattern obtained by this method can shorten the operating time compared to the conventionally known speed pattern.

[0101]

[Table 1]

[0102] From now on, by using the speed pattern of the present invention, the operating time is reduced by 17% compared to the case where acceleration and deceleration parameters are optimized using the cycloid speed pattern that has been commonly used in the past. can be shortened.

Next, a method of associating operating condition parameters and optimal acceleration/deceleration parameters using this method is shown in FIG.
This will be explained using the flowchart shown in . The above-mentioned method is used as is for the process of associating. By using this method, accurate optimum acceleration/deceleration parameters can always be determined regardless of operating condition parameters. However, there is a problem in that the amount of calculation is large and it takes a long time to install it in the control device of the drive device, or it requires a high-speed calculation processor.

Next, regarding the second embodiment, FIGS. 10 and 1
1 will be used for explanation. This embodiment is a method of establishing a relationship between the operating condition parameters obtained by the method described in the first embodiment and the optimum acceleration/deceleration parameters based on a table data utilization method. In this method, the calculations described in the first embodiment are not performed sequentially, but are used in the form of a data table, thereby significantly shortening the processing time. In this embodiment, a single-axis mechanism will be explained as an example.

FIG. 10 shows a flowchart of the table data utilization method, and FIG. 11 shows a table that relates operating condition parameters and optimum acceleration/deceleration parameters.

Each step in the flowchart shown in FIG. 10 will be explained.

(Step 1) Search for the closest operating condition parameter that is tabulated. Since the number of data to be tabulated is finite, the given operating condition parameter ( Hq, Δθq),
Select the closest data (Ht, Δθt).

When searching for the closest operating condition parameter, select the one that minimizes the squared error Δ2 shown in Equation 27.

[0109]

[Math. 27]

Here, w1 and w2 are weighting coefficients for adjusting the scale of the operating condition parameters H and Δθ. Tee
The bull is constructed as shown in FIG.

(Step 2) In step 1 of extracting the optimal acceleration/deceleration parameters, the vertical and horizontal axes to be referred to in the table were determined, so the optimal acceleration/deceleration parameters written in the table were determined.
Extract data.

(Step 3) Generation of speed pattern data Generate time history speed data or position data based on the optimal acceleration/deceleration parameters obtained in step 2, and issue speed commands to the drive device. Or send it as a position command.

As described above, by using the method described in this embodiment, it is possible to control the acceleration/deceleration of the drive device by a simple method of simply referring to table data.

However, high accuracy cannot be obtained unless a detailed table is prepared.

Next, a third embodiment of the present invention will be explained using FIGS. 12 and 13. This embodiment is a method of establishing a relationship between the operating condition parameters obtained by the method described in the first embodiment and acceleration/deceleration parameters based on a table data interpolation method. This method uses a table data
While data usage methods require detailed tables,
This was devised to achieve equivalent or higher accuracy by using a rougher table. In this embodiment, a single-axis mechanism will be described.

FIG. 12 shows a flowchart of the table data interpolation method, and FIG. 13 shows operating condition parameters, optimum acceleration/deceleration parameters, and operating condition parameter differential values of the optimum acceleration/deceleration parameters. This shows a table that relates the following.

Each step in the flowchart shown in FIG. 12 will be explained.

(Step 1) Search for the closest operating condition parameter in the table The same method as in the second embodiment is used. The table is shown in Figure 13.
It is made as shown.

(Step 2) In step 1 of extracting adjacent optimal acceleration/deceleration parameters and their differential values, the vertical and horizontal axes to be referred to in the table were determined, so the values written in the table Adjacent optimal acceleration/deceleration parameters and operating condition parameter differential value data of the optimal acceleration/deceleration parameters are extracted.

(Step 3) Calculation of interpolated values of optimal acceleration/deceleration parameters In order to find the optimal acceleration/deceleration parameters for a given operating condition, calculate the optimal acceleration/deceleration parameters for the adjacent operating conditions obtained in step 2. y, the operating condition parameter differential value dy/dx of the optimum acceleration/deceleration parameter, and the operating condition parameter error Δx.

[0121]

[Math. 28]

(Step 4) Generation of speed pattern data Generate time history speed data or position data based on the optimal acceleration/deceleration parameters obtained in step 3, and issue speed commands to the drive device. Or send it as a position command.

As described above, by using the method described in this embodiment, by referring to the table data and performing addition, subtraction, and multiplication, it becomes possible to control the acceleration/deceleration of the drive device using a simple method. . Compared to the table data utilization method described in the second embodiment, there is an advantage that the same or higher accuracy can be obtained by preparing a coarser table.

Next, a fourth embodiment of the present invention will be explained using FIGS. 14 and 15. This embodiment is a method in which the operating condition parameters obtained by the method described in the first embodiment are correlated with acceleration/deceleration parameters based on a least squares polynomial approximation method. In this embodiment, a single-axis mechanism will be described. FIG. 14 shows a procedure for deriving a least squares approximation polynomial, and FIG. 15 shows a flowchart of the least squares polynomial approximation method.

Each procedure in the flowchart shown in FIG. 14 will be explained.

(Procedure 1) Determining a polynomial The optimum acceleration/deceleration parameters are related to the operating condition parameters using a polynomial as shown in Equation 29, and the order of terms to be determined is determined.

[0127]

[Math. 29]

(Step 2) Determination of polynomial coefficients Polynomial coefficients are determined so as to minimize the least squares error shown in Equation 30.

[0129]

[Math. 30]

FIG. 15 shows a flowchart for generating a speed pattern of a drive device using this method.

[0131] Each step will be explained.

(Step 1) Calculation of optimal acceleration/deceleration parameters based on approximate polynomials Calculate optimal acceleration/deceleration parameters for the given operating condition parameters using equation 29 with polynomial coefficients calculated and stored in advance. Calculate parameters.

(Step 2) Generation of speed pattern data Generate time history speed data or position data based on the optimal acceleration/deceleration parameters obtained in step 1, and issue speed commands to the drive device. Or send it as a position command.

As described above, by using the method described in this embodiment, it is possible to control the acceleration/deceleration of the drive device using a simple method using only the polynomial calculation shown in Equation 29.

[0135] This method uses table data usage method,
This enables calculation of acceleration/deceleration parameters with higher precision than the bull data interpolation method. However, there are problems in that selecting the number of terms in the polynomial shown in Equation 29 requires some trial and error, and deriving the polynomial coefficient calculation formula shown in Equation 30 becomes somewhat complicated as the number of terms increases.

Next, regarding the fifth embodiment of the present invention, FIG.
This will be explained using FIGS. 6 to 18.

[0137] This embodiment is a method of establishing a relationship between the operating condition parameters obtained by the method described in the first embodiment and acceleration/deceleration parameters based on a neural network learning method. . This embodiment uses a simple learning method that does not involve deriving equations, compared to the least squares polynomial approximation method described in the fourth embodiment. This embodiment targets a single-axis mechanism.

FIG. 16 shows the structure of the neural network, FIG. 17 shows a flowchart of the neural network learning method, and FIG. 18 shows the convergence status of acceleration/deceleration parameter errors during learning. It shows.

[0139] In the learning stage, the menu shown in Fig.
The paired data obtained by the method described in the first embodiment is given to the input layer and output layer of the network as teacher data, and the weighting coefficient between each layer is determined. Here, among the acceleration/deceleration parameters, it is known that the acceleration index and deceleration index are optimally 1.0 and 1.2, respectively.
Output parameters for the other three acceleration/deceleration parameters
treated as data. After determining the weighting coefficient between layers, Figure 17
The speed pattern is generated according to the flowchart shown in . Each process will be explained next.

(Step 1) Calculation of optimal acceleration/deceleration parameters using the weighting coefficients obtained during learning Using the weighting coefficients between each layer obtained during the learning stage, the equation 31 is calculated.
is used to determine the optimal acceleration/deceleration parameters for the given operating condition parameters.

[0141]

[Math. 31]

[0142] f is a sigmoid function.

(Step 2) Generation of speed pattern data Generate time history speed data or position data based on the optimal acceleration/deceleration parameters obtained in step 1, and issue speed commands to the drive device. Or send it as a position command.

Next, the calculation results will be shown. First, Figure 1 shows the convergence status of acceleration/deceleration parameter errors when learning was performed by giving 16 pieces of learning data to the single-axis mechanism used in the first embodiment.
8. From this, sufficient convergence can be seen for each parameter after about 10,000 learning times.

Next, Table 2 shows the results of determining acceleration/deceleration patterns in accordance with the flowchart of FIG. 17 under various operating conditions.

[0146]

[Table 2]

[0147] From now on, when the operating condition parameters and the optimum acceleration/deceleration parameters are related using a neural network, the above results show an error of about 10% at most. In addition, there are cases where the torque margin becomes negative, and this correlation error becomes a bottleneck when using the system. this is,
Improvements can be made by increasing the number of learning points, increasing the number of hidden layer elements in the neural network, increasing the number of layers, etc.

Next, a sixth embodiment of the present invention will be explained using FIG. 19. This embodiment is intended to improve the accuracy of the least squares polynomial approximation method described in the fourth embodiment and the neural network learning method described in the fifth embodiment. FIG. 19 shows a plurality of subdivided regions for approximation and learning.

[0149] In order to obtain highly accurate optimal acceleration/deceleration parameters in a wide range of operating condition parameters using the least squares polynomial approximation method and neural network learning method, it is necessary to If the data range is selected to one digit as shown in area A in FIG. 19, sufficient accuracy cannot be obtained. Therefore, by dividing the area into a plurality of areas as shown in areas A1, A2, and A3, the approximation and learning accuracy for each area can be improved, and the accuracy as a whole can be improved.

Next, a seventh embodiment of the present invention will be explained using FIG. 20. This embodiment is designed to prevent the drive device from generating torque shortage due to the error in the relationship between the operating condition parameters and the optimum acceleration/deceleration parameters described in the second, third, fourth, and fifth embodiments. This is an improved version. FIG. 20 is a diagram showing a procedure for determining optimal acceleration/deceleration parameters for operating condition parameters.

[0151] In this embodiment, in order to prevent the drive device from running out of torque due to relationship errors between parameters as seen in Table 2, the optimum acceleration/deceleration parameters are set taking into consideration the margin. It is a method of making decisions. FIG. 7 shows the procedure, and shows the procedure corresponding to FIG. 2 of the first embodiment.

Next, an eighth embodiment of the present invention will be explained using FIG. 21. This embodiment specifically focuses on the load inertia motor among the operating condition parameters used to determine the optimal acceleration/deceleration parameters.
This method aims to improve the accuracy of the optimum acceleration/deceleration parameters obtained by accurately identifying the element H. FIG. 21 shows the configuration of a control device for a drive device including a torque detection means for the drive device. In FIG. 21, the torque detection means 7 is comprised of a shunt resistor for current detection, etc. The torque detection means 7 detects the required torque when the drive device performs the reference operation, and the load inertia moment can be determined under actual operating conditions by using the equation of motion based on the result.

Next, an application example of the acceleration/deceleration control method for a drive device according to the present invention will be described with reference to FIGS. 22 to 25. Although the above-mentioned first to eighth embodiments have been described with reference to a single-axis mechanism, here, the structure of a multi-axis mechanism such as a robot and a robot system to which the present method is applicable will be described.

FIG. 22 shows an external view of the horizontal articulated robot, FIG. 23 shows the relationship between the parameters of the two-axis mechanism, and FIG. 24 shows an external view of the vertical articulated robot.
FIG. 25 shows the relationship between the parameters of the three-axis mechanism.

[0155] The horizontal articulated robot (see Fig. 22) is
A first arm 13 is rotationally driven by a first drive device 12 installed on the base, and a second arm is rotationally driven by a second drive device 14 attached to the tip of the first arm. -mu1
5, and a second drive unit driven in the vertical direction and rotational direction by the third and fourth drive devices 16 and 17 via the power transmission member.
It is composed of a wrist shaft 18 provided at the tip of the arm. Positioning of the tool 19 provided at the tip of the wrist shaft 18 in a horizontal plane is realized by first and second drive devices,
Vertical positioning is achieved by a third drive. Further, the fourth drive device determines the posture of the wrist axis.

[0156] To realize work that requires high-speed movement on a horizontal work surface, such as assembly work, each axis of the two-axis mechanism consisting of the first and second drive devices has an acceleration/deceleration pattern (10 patterns in total). Optimization is required (see Figure 23). By implementing the correlation function section in FIG. 23 using a neural network or the like, the operating time can be minimized.

[0157] The vertical articulated robot (see Fig. 24) is
A swivel base 21 rotatably driven by a swivel shaft drive device 20 installed on the base, and an upper arm shaft drive device 23 that rotatably drives the upper arm arm 22 on the swivel base 21 in a direction orthogonal to the pivot axis.
, a forearm shaft drive device 25 that rotationally drives the forearm arm 24 provided at the tip of the upper arm arm, and a fourth, fifth, and sixth
The wrist part 26 is provided at the tip of a forearm arm that is driven in rotation, bending, and twisting directions by a driving device. In a vertically articulated robot, movement of the wrist between two points in a three-dimensional plane is realized by the rotation axis, upper arm axis, and forearm axis drive devices, and the posture of the wrist is determined by the fourth, fifth,
and a sixth driving device. Therefore, in order to realize work requiring high-speed operation, optimization of the acceleration/deceleration parameters (15) for each axis of the three-axis mechanism is required (see FIG. 25).

[0158] The robot system consisting of the above-mentioned robot main body and a control device with a related function section can realize high-speed operation.

[0159]

[Effects of the Invention] Since the present invention is constructed as described above, it produces the effects as described below.

(1) The speed pattern of the drive device is generated using an exponential sine wave curve, which is a highly flexible acceleration/deceleration curve with a large number of acceleration/deceleration parameters. It becomes possible to obtain an optimal speed pattern.

(2) Dynamic characteristics of the moving body and drive motor
The optimum acceleration/deceleration parameters are determined to minimize the driving force margin by taking into consideration the maximum generated driving force vs. speed characteristics of the motor, making it possible to perform acceleration/deceleration control that fully utilizes the performance of the drive device. Become.

(3) The relationship between operating condition parameters (operating distance, load body inertia moment, etc.) and optimal acceleration/deceleration parameters (acceleration/deceleration index, maximum speed, specific acceleration/deceleration distance) is expressed as (i
) table data usage method, (ii) table data interpolation method, (iii) least squares polynomial approximation method, (iv)
Since this is done using a neural network learning method, the optimum acceleration/deceleration parameters can be easily extracted for the given operating conditions. (4) Since the operating condition parameter identification process is provided, it is possible to calculate the optimum acceleration/deceleration parameters with high accuracy.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a speed pattern of a drive device of the present invention.

FIG. 2 is a speed pattern optimization procedure diagram.

FIG. 3 is an external view of a single-axis mechanism and its control device.

FIG. 4 shows convergence characteristics of operating time.

FIG. 5 shows convergence characteristics of acceleration/deceleration index.

FIG. 6 shows the convergence characteristics of the maximum speed.

FIG. 7 is a convergence characteristic of specific acceleration/deceleration distance.

FIG. 8 is a comparison of operating time with a conventional acceleration/deceleration curve.

FIG. 9 is a flowchart for determining optimal acceleration/deceleration parameters.

FIG. 10 is a flowchart of a table data usage method.

FIG. 11 is a table of a table data usage method.

FIG. 12 is a flowchart of a table data interpolation method.

FIG. 13 is a table of the table data interpolation method.

FIG. 14 is a procedure for deriving a least squares approximation polynomial.

FIG. 15 is a flowchart of a least squares polynomial approximation method.

FIG. 16 shows the structure of a neural network.

FIG. 17 is a flowchart of a neural network learning method.

FIG. 18 shows convergence characteristics of parameter errors during learning.

FIG. 19 is an approximation/learning area diagram of operating condition parameters.

FIG. 20 is a procedure for calculating optimal acceleration/deceleration parameters in consideration of motor torque margin.

FIG. 21 is a control device for a drive device having a torque detection means.

FIG. 22 is an external view of a horizontal articulated robot.

FIG. 23 is a relationship between parameters of a two-axis mechanism.

FIG. 24 is an external view of a vertically articulated robot.

FIG. 25 shows the relationship between parameters of the three-axis mechanism.

FIG. 26 shows the operation of a high-speed moving body.

FIG. 27 is a speed pattern of a high-speed moving object.

FIG. 28 is an exponential sine wave curve.

FIG. 29 shows temporal changes in dimensionless parameters.

FIG. 30 shows the relationship between the dimensionless parameter maximum value and the index.

FIG. 31 shows the relationship between maximum acceleration time and index.

FIG. 32 is an external view of the multi-axis mechanism.

FIG. 33 shows the generated torque-speed characteristics of the motor.

FIG. 34 shows the change in required torque over time under optimal driving conditions.

FIG. 35 shows the configuration of a neural network.

FIG. 36 shows convergence characteristics of parameter errors.

[Explanation of symbols]

Claims (11)

[Claims] Claim 1: In an acceleration/deceleration control method of a drive device, a speed pattern θ of a drive shaft during a moving operation is selected as shown in equation 1 using an exponential sine wave curve, [Equation 1] ■ Acceleration index m1 ,■Deceleration index m2,■Maximum speed θm,■
A speed pattern is planned using specific acceleration distance α (ratio of acceleration distance to operating distance) and ■ specific deceleration distance β (ratio of deceleration distance to operating distance) as (variable) acceleration/deceleration parameters, and the drive device control device A method for controlling acceleration/deceleration of a drive device, characterized in that the acceleration/deceleration control method is provided as a speed command to 2. The acceleration/deceleration control method for a drive device according to claim 1, wherein the acceleration index is selected such that m1≧1 and the deceleration index is selected such that m2>1. . 3. The acceleration/deceleration control method for a drive device according to claim 1 or 2, wherein given operating conditions (i.e. the relationship between the generated driving force and the speed of the drive device; and - The driving force margin of the drive device in the acceleration section, constant velocity section, and deceleration section is the minimum with respect to the 1. A method for controlling acceleration/deceleration of a drive device, characterized in that acceleration/deceleration parameters are determined so as to satisfy the following. 4. The acceleration/deceleration control method for a drive device according to claim 1 or 2, wherein the relationship between the optimum acceleration/deceleration parameter with respect to the plurality of operating condition parameters obtained by the method according to claim 3 is tested. - A method for controlling acceleration/deceleration of a drive device, characterized in that the optimum acceleration/deceleration parameters for arbitrary operating conditions are obtained by converting table data close to a given operating condition into table data. 5. The acceleration/deceleration control method for a drive device according to claim 1 or 2, wherein the optimum acceleration/deceleration parameter and the optimum acceleration/deceleration parameter for the operating condition parameters obtained by the method according to claim 3 are provided. The relationship between the operating condition parameter differential values of
1. An acceleration/deceleration control method for a drive device, characterized in that an optimum acceleration/deceleration parameter for a given operating condition is determined from bull data using an interpolation formula. 6. The acceleration/deceleration control method for a drive device according to claim 1 or 2, wherein a plurality of (operating condition parameters, optimal acceleration/deceleration parameters) obtained by the method according to claim 3 are provided.
By giving this data to the input/output layer of the neural network as training data and letting it learn, the weighting coefficients between the input layer/hidden layer and the hidden layer/output layer are determined, and the weighting coefficients with the weighting coefficients are determined. 1. An acceleration/deceleration control method for a drive device, characterized in that, based on a neural network, optimal acceleration/deceleration parameters for arbitrary operating conditions are determined and a speed pattern is generated. 7. The acceleration/deceleration control method for a drive device according to claim 1 or 2, wherein a plurality of (operating condition parameters, optimal acceleration/deceleration parameters) obtained by the method according to claim 3 are provided.
1. An acceleration/deceleration control method for a drive device, characterized in that the data of the above are related using an approximation polynomial, and the coefficients thereof are determined by the least squares method. 8. The acceleration/deceleration control method for a drive device according to claim 6 or 7, wherein the operating condition is divided into a plurality of regions and associated with each other. 9. In the acceleration/deceleration control method for a drive device according to claim 1, in determining the optimum acceleration/deceleration parameter according to claim 3, table data interpolation error, least squares polynomial approximation error, or 1. An acceleration/deceleration control method for a drive device, characterized in that the drive device is determined to have a drive force margin equivalent to a learning error of the drive device. 10. The acceleration/deceleration control method for a drive device according to claim 1, further comprising the step of identifying a load inertia moment of the drive device. 11. A robot system in which a movable body having a working part at its tip is moved using power generated by a drive device to perform a positioning operation, wherein the control device of the drive device is provided with the addition according to any one of claims 1 to 10. A robot system characterized in that it is driven by giving a speed pattern generated using a deceleration control method or a position pattern that is an integral amount thereof as a speed command or a position command.