Observer-corrector control system for systems with unmodeled dynamics

• US 6,567,711 B1
• Filed: 08/28/2000
• Issued: 05/20/2003
• Est. Priority Date: 08/28/2000
• Status: Expired due to Term

• Current Assignee: Brooks Automation Incorporated (Azenta, Inc.)
• Original Assignee: Brooks Automation Incorporated (Azenta, Inc.)
• Inventors: Hosek, Martin, Moura, Jairo Terra, Elmali, Hakan
• Primary Examiner(s): Patel, Ramesh
• Application Number:
  US09/649,374
• Time in Patent Office: 995 Days
• Field of Search: 700/55, 700/30, 700/33, 700/34, 700/38, 700/37, 700/39, 700/45-46, 180/197, 701/4, 701/37, 701/70, 701/72, 701/110, 701/125, 701/29, 381/98, 381/61, 381/106, 381/119
• US Class Current: 700/55
• CPC Class Codes: G05B 13/021 in which a variable is automatically adjusted to optimise the performance

Abstract

A system for extracting a signal component from an output signal of a dynamic system. The system comprises a state observer and a corrector filter. The state observer is adapted to track a signal component that represents dominant dynamics in the output signal of the dynamic system and provide an estimation signal representing an estimated signal component in the output signal of the dynamic system. The corrector filter is adapted to compensate for a mismatch between the estimation signal and the actual signal component that represents the dominant dynamics in the output signal. A combination of the estimation signal with an output signal of the corrector filter can provide a synthesized signal including the signal component that represents the dominant dynamics in the output signal.

Claims

1. A system for extracting a signal component that represents dominant dynamics in an output signal of a dynamic system comprising:

a state observer adapted to track an actual signal component that represents dominant dynamics in the output signal of the dynamic system, and provide an estimation signal representing an estimated signal component that represents dominant dynamics in the output signal of the dynamic system; and

a corrector filter adapted to compensate for a mismatch between the estimation signal and the actual signal component that represents dominant dynamics in the output signal;

wherein a combination of the estimation signal with an output signal of the corrector filter can provide a synthesized signal including the signal component that represents dominant dynamics in the output signal.

2. The system of claim 1 further comprising a controller for the dynamic system, the synthesized signal forming a feedback signal used as an input to the controller, wherein a destabilizing effect of unmodeled higher order dynamics in the output signal of the dynamic system is reduced or substantially eliminated.

3. The system of claim 2 wherein the system is a control system for a machine.

4. The system of claim 2 wherein the state observer is adapted to receive a first input representing an output signal from the controller and a second input representing the dynamic system output signal.

5. The system of claim 4 wherein the second input representing the dynamic system output signal includes a measurement noise signal component.

6. The system of claim 2, wherein a design of the controller is based upon a model of the dynamic system that does not include the higher-order dynamics of the dynamic system.

7. The system of claim 1 wherein the dynamic system is a robotic manipulator.

8. The system of claim 1 wherein the synthesized signal does not include a signal component that represents higher order dynamics in the dynamic system output signal.

9. The system of claim 1 wherein the state observer is further adapted to track the actual signal component that represents dominant dynamics in the output signal of the dynamic system and not respond to a higher-frequency signal component of the output signal of the dynamic system, the higher-frequency signal component corresponding to a higher-order dynamics of the dynamic system and a measurement noise signal component of the dynamic system output signal.

10. The system of claim 1 wherein the state observer is a band-limited state observer.

11. The system of claim 1 wherein the corrector filter is further adapted to pass all lower-frequency signal components associated with the actual signal component that represents dominant dynamics together with an observation error signal component and attenuate all higher-frequency signal components associated with the higher order dynamics and a measurement noise signal component.

12. The system of claim 1 wherein the corrector filter is a linear low-pass filter.

13. A method of extracting a signal component from an output signal of a dynamic system comprising the steps of:

estimating a signal component that represents dominant dynamics in the dynamic system output signal;

compensating for a mismatch between the estimated signal component and an actual signal component that represents dominant dynamics in the dynamic system output signal;

combining the estimated signal component with a signal representing the compensation for a mismatch between the estimated signal component and the actual signal component to provide a synthesized signal including the signal component that represents dominant dynamics in the output signal.

14. The method of claim 13 wherein the extracted signal component is a dominant signal component of the output signal.

15. The method of claim 13 further comprising the step of using the synthesized signal as a feedback input signal for a controller for the dynamic system, wherein a destabilizing effect of unmodeled higher order dynamic signal components in the dynamic system output signal are reduced or substantially eliminated.

16. The method of claim 15, wherein the step of estimating a signal component that represents dominant dynamics in the dynamic system output signal comprises the step of tracking the actual signal component that represents dominant dynamics in the dynamic system output signal while excluding a higher order signal component and a measurement noise signal component in the dynamic system output signal.

17. The method of claim 15 wherein the step of compensating for a mismatch between the estimated signal component and the actual signal component that represents dominant dynamics in the dynamic system output signal comprises the step of passing a lower-frequency signal component that represents dominant dynamics in the dynamic system output signal and an observation error while attenuating a higher-order frequency signal component and a measurement noise signal component of the dynamic system output signal.

18. A method of reducing effects of higher order dynamics in a controlled system comprising the steps of:

tracking an actual signal component that represents dominant dynamics in an output signal of the controlled system;

providing an estimation signal representing an estimated component that represents dominant dynamics in the output signal of the controlled system;

compensating for a mismatch between the estimation signal and the actual signal component that represents dominant dynamics;

combining the estimation signal with an output signal from the corrector filter to form a synthesized feedback signal, the synthesized feedback signal including a signal component that represents dominant dynamics in the output signal of the controlled system; and

inputting the synthesized feedback signal to a controller for the controlled system, wherein a destabilizing effect of unmodeled higher order dynamic signal components in the dynamic system output signal is reduced or substantially eliminated.

19. The method of claim 18 wherein the controlled system is a robotic manipulator and the step of tracking an actual signal component that represents dominant dynamics in the output signal of the controlled system comprises tracking rigid body dynamics of the manipulator within a specified bandwidth.

20. The method of claim 19 wherein the step of combining the estimation signal with an output signal from the corrector filter to form a synthesized feedback signal comprises the step of combining an observed velocity signal of a motor with an output of the corrective filter to yield a clean synthetic signal which closely follows a raw velocity signal of the motor but does not contain the undesirable high-frequency components contained in the raw velocity signal.