SYSTEM AND METHOD FOR ADAPTIVE CONTROL OF UNCERTAIN NONLINEAR PROCESSES

• US 20070010902A1
• Filed: 06/07/2005
• Published: 01/11/2007
• Est. Priority Date: 08/01/1995
• Status: Active Grant

• Current Assignee: Guided Systems Technologies, Inc.
• Original Assignee: Guided Systems Technologies, Inc.
• Inventors: Kim, Byoung-Soo, Calise, Anthony
• Granted Patent: US 7,177,710 B2
• Time in Patent Office: Days
• US Class Current: 700/51
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G05D 1/0825 using mathematical models

Abstract

A computer system for controlling a nonlinear physical process. The computer system comprises a linear controller and a neural network. The linear controller receives a command signal for control of the nonlinear physical process and a measured output signal from the output of the nonlinear physical process. The linear controller generates a control signal based on the command signal, a measured output signal, and a fixed linear model for the process. The neural network receives the control signal from the linear controller and the measured output signal from the output of the nonlinear physical process. The neural network uses the measured output signal to modify the connection weights of the neural network. The neural network also generates a modified control signal supplied to the linear controller to iterate a fixed point solution for the modified control signal used to control the nonlinear physical process.

Claims

1. A computer system for controlling a non-linear physical process, the computer system comprising:

a linear controller connected to receive a command signal for control of the non-linear physical process and an output signal from the non-linear physical process, the linear controller generating a control signal based on the command signal and the output signal, the linear controller designed using a fixed linear model for the process; and

a neural network connected to receive a combined control signal and the output signal from the non-linear physical process, the neural network using at least one of the combined control signal and the output signal from the non-linear physical process to modify the connection weights of the neural network on-line as the neural network and linear controller are used to control the non-linear physical process, the neural network generating a modified control signal based on the modified connection weights and at least one of the combined control signal and the output signal from the non-linear physical process, the modified control signal combining with the control signal to form the combined control signal used to control the non-linear physical process and to correct for errors inherent in controlling the physical process using the linear controller designed using the fixed linear model.

2. A computer system as claimed in claim 1 wherein the linear controller is a proportional-derivative (PD) controller.

3. A computer system as claimed in claim 1 wherein the linear controller is a proportional-integral-derivative (PID) controller.

4. A computer system as claimed in claim 1 wherein the neural network is trained entirely on-line based on the output signal.

5. A computer system as claimed in claim 1 wherein the linear controller comprises a plurality of attitude controllers and the neural network comprises a plurality of neural network subunits, the attitude controllers and neural network subunits generating respective degrees of freedom of the modified control signal.

6. A computer system as claimed in claim 1 further comprising:

a command transformation unit connected to receive a body angular rate signal, and generating the command signal based on the body angular rate signal.

7. A computer system as claimed in claim 6 wherein the command signal includes the first derivatives of commanded roll, pitch, and yaw attitudes of an aircraft.

8. A computer system as claimed in claim 7 further comprising:

an integrator connected to receive the first derivatives of the commanded roll, pitch, and yaw attitude signals, and integrating the first derivatives of the commanded roll, pitch, and yaw attitude signals to generate the commanded roll, pitch, and yaw signals, the integrator connected to supply the commanded roll, pitch, and yaw signals to the linear controller as a part of the command signal.

9. A computer system as claimed in claim 6 further comprising:

a command augmentation unit connected to receive roll rate and acceleration command signals from a pilot, the command augmentation unit generating the body angular rate signal based on the roll rate and acceleration command signals from the pilot.

10. A computer system as claimed in claim 9 wherein the command augmentation unit further generates the body angular rate signal based on side slip in addition to the roll rate and acceleration command signals from the pilot.

11. A computer system as claimed in claim 1 further comprising:

an inverse function unit connected to receive the combined control signal and modified control signal, the inverse function unit generating an actuator control signal for controlling the non-linear physical process based on the combined control signal and modified control signal.

12. A computer system as claimed in claim 1 further comprising:

an inertial measurement unit (IMU) connected to sense the output of the non-linear physical process, the IMU generating the output signal based on the output of the non-linear physical process.

13. A computer system as claimed in claim 1 wherein the non-linear physical process relates to flight control of an aircraft.

14. The computer system as claimed in claim 1 wherein the linear controller and the neural network are used to control the non-linear physical process using a moving surface controlled by a hydraulic actuator based on the signals generated by the linear controller and the neural network.

15. The computer system as claimed in claim 1 wherein the linear controller and the neural network are used to control the non-linear physical process using a moving surface controlled by an electrical actuator based on the signals generated by the linear controller and the neural network.

16. The computer system as claimed in claim 1 wherein the linear controller and the neural network are used to control the non-linear physical process by controlling airflow over the surface of an aircraft using synthetic jets based on the signals generated by the linear controller and the neural network.

17. A computer system for controlling a non-linear physical process, the computer system comprising:

a linear controller connected to receive a command signal for control of the non-linear physical process and an output signal from the non-linear physical process, the linear controller generating a control signal based on the command signal and the output signal, the linear controller designed using a fixed linear model for the process;

a neural network connected to receive a combined control signal and the output signal from the non-linear physical process, the neural network using at least one of the combined control signal and the output signal from the non-linear physical process to modify the connection weights of the neural network on-line as the neural network and linear controller are used to control the non-linear physical process, the neural network generating a modified control signal based on the modified connection weights and at least one of the combined control signal and the output signal from the non-linear physical process; and

a node connected to receive the control signal output from the linear controller and the modified control signal output from the neural network, and generating the combined control signal based on the control signal output from the linear controller and the modified control signal output from the neural network, the node connected to output the combined control signal to control the non-linear physical process and to correct for errors inherent in modeling the physical process using the fixed linear model.

18. A computer system for controlling a non-linear physical process, the computer system comprising:

a controller connected to receive a command signal for control of the non-linear physical process and an output signal from the non-linear physical process, the controller generating a control signal based on the command signal and the output signal, the controller designed using a model for the process;

a neural network connected to receive a combined control signal and the output signal from the non-linear physical process, the neural network using at least one of the combined control signal and the output signal from the non-linear physical process to modify the connection weights of the neural network on-line as the neural network and controller are used to control the non-linear physical process, the neural network generating a modified control signal based on the modified connection weights and at least one of the combined control signal and the output signal from the non-linear physical process, the modified control signal combining with the control signal to form the combined control signal used to control the non-linear physical process and to correct for errors inherent in modeling the physical process using the model.

19. A computer system for controlling a non-linear physical process, the computer system comprising:

a controller connected to receive a command signal for control of the non-linear physical process and an output signal from the non-linear physical process, the controller generating a control signal based on the command signal and the output signal, the controller designed using a model for the process;

a neural network connected to receive a combined control signal and the output signal from the non-linear physical process, the neural network using at least one of the combined control signal and the output signal from the non-linear physical process to modify the connection weights of the neural network on-line as the neural network and controller are used to control the non-linear physical process, the neural network generating a modified control signal based on the modified connection weights and at least one of the combined control signal and the output signal from the non-linear physical process; and

a node connected to receive the control signal output from the controller and the modified control signal output from the neural network, and generating the combined control signal based on the control signal output from the controller and the modified control signal output from the neural network, the node connected to output the combined control signal to control the non-linear physical process and to correct for errors inherent in modeling the physical process using the model.