Prediction controller

Details Prediction controller Prediction controller

1995-12-26

1998-03-10

Elmore, Reba I.

Boots, shoes, and leggings

364132, 364187, 395187, G05B 1302

Patent

active

057268789

DESCRIPTION:

BRIEF SUMMARY
The present invention relates to a controller for machine tools, robots, etc.
The applicant has proposed a prediction control technique in Japanese unexamined patent publication No. Hei-3-203882 (Japanese examined patent publication No. Hei-5-27829). According to this technique, the first several step responses of an object to be controlled are sampled and control inputs are determined so that a predicted deviation is minimized, using a model which is an approximation obtained on the assumption that the increment will be attenuated at a constant ratio.
Prediction control techniques utilizing a transfer function model of an object to be controlled include that described by Hashimoto, Kuroyanagi, and Harashima in "Prediction Control of Servo Systems Utilizing DSP", p. 990-996 of Bulletin D No. 9, the Institute of Electrical Engineers, 1990, and that described by Tamura, Nishitani, and Kunugida in "Model Prediction Control of Integral Processes" p. 367, Vol. 56 of the Memoir of Annual Lecture Meeting of Chemical Industries Association, 1991.
The technique disclosed in Japanese unexamined patent publication No. Hei-3-203882 requires step responses from an object to be controlled to predict future deviations. The step responses may be calculated by means of simulation if a transfer function model of the object to be controlled is available. However, this process is indirect and involves corresponding time and labor.
The technique proposed by Hashimoto et al. wherein a transfer function model is directly used to predict a deviation has had problems in that a great deal of calculation is involved in measuring the control input performed in each sampling period; since the performance criterion does not include the term of control input, the range for prediction must be large to keep the degree of correction low, which further increases the amount of calculation; since a target command, output, and control input are used as they are, the data has a large bit length which requires a large memory and significant calculation; and there is no way to deal with a delay in the detection of the output to be controlled.
The technique proposed by Tamura et al. has had problems in that it involves matrix calculations of higher orders, including inverse matrix calculations, when the model is at a high order or the range for prediction is large; since a target command, output, and control input are used as they are, the data has a large bit length which requires a large memory and a lot of calculation; and there is no way to deal with a delay in the detection of the output to be controlled.


SUMMARY OF THE INVENTION

It is an object of the present invention to provide a predictive controller which requires only a small amount of memory and minimal calculation during each sampling period, and which can cope with a delay in the detection of the output to be controlled.
In a prediction controller wherein the output of an object to be controlled has a transfer function model of (1-a.sub.1 z.sup.-1 - . . . -a.sub.Na z.sup.-Na)} increment .DELTA.r(i+M) for the target command at the M-th sampling in the future and an increment .DELTA.y(i-K) as an output from the object to be controlled at the K-th sampling (K.gtoreq.0) in the past representation of an actual displacement and by outputting a control input v(i) to the object to be controlled at the current point in time i, characterized in that it includes a means for storing the increment for the target command, constants for prediction control, the increment for the past output, and control input in the past, a means for obtaining a deviation e(i-K) from the increment for the target command and the increment for the output of the object to be controlled, and a means for determining the control input v(i) at the current point in time from: ##EQU1## where vm, pn, E, and gn are constants for prediction control.
The above-described means make it possible to provide a prediction controller which requires only a small amount of memory and minimal calculation during each sampling period and

REFERENCES:
patent: 4615723 (1986-10-01), Rodriguez-Fernandez et al.
patent: 4931879 (1990-06-01), Koga et al.
patent: 5369345 (1994-11-01), Phan et al.
patent: 5463746 (1995-10-01), Brodnax et al.

Affiliated with

Also associated with

Rating

Prediction controller does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Prediction controller, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Prediction controller will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-144641