초록 ▼

A process and neural network architecture for on-line adjustment of the weights of the neural network in a manner that corrects errors made by a nonlinear controller designed based on a model for the dynamics of a process under control. A computer system is provided for controlling the dynamic outpu

A process and neural network architecture for on-line adjustment of the weights of the neural network in a manner that corrects errors made by a nonlinear controller designed based on a model for the dynamics of a process under control. A computer system is provided for controlling the dynamic output response signal of a nonlinear physical process, where the physical process is represented by a fixed model of the process. The computer system includes a controlled device for responding to the output response signal of the system. The computer system also includes a linear controller for providing a pseudo control signal that is based on the fixed model for the process and provides a second controller, connected to the linear controller, for receiving the pseudo control signal and for providing a modified pseudo control signal to correct for the errors made in modeling the nonlinearities in the process. A response network is also included as part of the computer system. The response network receives the modified pseudo control signal and provides the output response signal to the controlled device. The second controller preferably is a neural network. The computer system may include a plurality of neural networks with each neural network designated to control a selected variable or degree of freedom within the system.

대표청구항 ▼

1. A computer system for controlling a nonlinear physical process with a dynamic output response signal, said physical process represented by a fixed model of said process, comprising:a linear controller for providing a pseudo control signal, said pseudo control signal being based on said fixed mode

1. A computer system for controlling a nonlinear physical process with a dynamic output response signal, said physical process represented by a fixed model of said process, comprising:a linear controller for providing a pseudo control signal, said pseudo control signal being based on said fixed model for said process;a second adaptive controller, connected to said linear controller, for receiving said pseudo control signal and for providing a modified pseudo control signal to said linear controller to correct for errors inherent in modeling of said process, said second adaptive controller comprising a neural network for modifying said pseudo control signal based on on-line analysis of said process, the value of said modified pseudo control signal equals the sum of said pseudo control signal, outputted from said linear controller, and an estimated derivative of a received command signal for a degree of freedom in said system, the sum being subtracted by an output signal from said neural network for that same degree of freedom; anda response network, connected to said second controller, for receiving said modified pseudo control signal and for providing said output response signal to a controlled device. 2. The computer system of claim 1 wherein the controlled device comprises a moving surface controlled by a hydraulic actuator based on the output response signal. 3. The computer system of claim 1 wherein the controlled device comprises a moving surface controlled by an electrical actuator based on the output response signal. 4. The computer system of claim 1 wherein the controlled device comprises a surface receiving airflow from a plurality of synthetic jets based on the output response signal. 5. A method for providing a control signal to a controlled device in a modeled dynamic non-linear process in which data is received on-line for a non-linear system, comprising the steps of:receiving feedback state and command signals at a controller;calculating a pseudo control signal for the received feedback state and command signals;receiving the pseudo control signal and state signal at an online neural network that is adapted to correct for errors inherent in the modeling of the non-linear physical process;calculating a fixed point solution at the neural network to ensure stability of the process;adjusting connection weights of the neural network based on the state and command signals received;modifying, at the neural network, the pseudo control signal with an output signal of the on-line neural network to correct for inverse modeling errors;producing a modified pseudo control signal as an output from said neural network;receiving the modified pseudo control signal and calculating an inverse response control signal at an inverse response function unit that is based on a model for the process; andtransmitting the inverse response control signal to an output response device for producing adjustments to the controlled device. 6. The method of claim 5 wherein the output response device comprises a hydraulic actuator device for moving the controlled device that comprises a moving surface of an aircraft based on the output response signal. 7. The method of claim 5 comprising the step of calculating said modified pseudo control signal by adding an output signal of said controller and a received control signal for a degree of freedom in said system to form a sum signal, and subtracting said sum signal by an output signal of said neural network. 8. The method of claim 5 wherein the output response device comprises an electrical actuator for moving a controlled device that comprises a surface of an aircraft based on the output response signal. 9. The method of claim 5 wherein the output response device comprises a plurality of synthetic jets for modifying airflow over a controlled device that comprises a surface of an aircraft based on the output response signal. 10. A computer system for controlling the dynamic output response of a nonline ar physical process, said physical process represented by a fixed model of said process, comprising:a linear controller for providing a pseudo control signal, said pseudo control signal being based on said fixed model for said process;a second controller, connected to said linear controller, for receiving said pseudo control signal and for providing a modified pseudo control signal to correct for errors inherent in modeling of said process, said second controller comprising a neural network for modifying said pseudo control signal based on on-line data training of said neural network, the value of said modified pseudo control signal equals the sum of value of the output of said linear controller and the value of an estimated derivative of a received command signal for a degree of freedom in said system subtracted by the value of the output of said neural network for that same degree of freedom; anda response network, connected to said second controller, for receiving said modified pseudo control signal and for providing an output response signal to a controlled device. 11. The computer system of claim 10 wherein the controlled device comprises a moving surface controlled by a hydraulic actuator based on the output response signal. 12. The computer system of claim 10 wherein the controlled device comprises a moving surface controlled by an electrical actuator based on the output response signal. 13. The computer system of claim 10 wherein the controlled device comprises a surface receiving a modified airflow from a plurality of synthetic jets based on the output response signal. 14. A method for providing a control signal to an actuator device for controlling an aircraft in a modeled dynamic non-linear process in which data is received on-line for a non-linear system, comprising the steps of:receiving feedback state and command signals at a controller;calculating a pseudo control signal for the received feedback state and command signals;receiving the pseudo control signal and state signal at an online neural network that is adapted to correct for errors inherent in the modeling of the non-linear physical process;calculating a fixed point solution at the neural network to ensure stability of the process;adjusting connection weights of the neural network based on the state and command signals received;modifying, at the neural network, the pseudo control signal with the output of the on-line neural network to correct for inverse modeling errors;producing a modified pseudo control signal as an output from said neural network by adding the value of the output of said linear controller and the value of a received control signal for a degree of freedom in said system to form a sum, and subtracting said sum by the value of the output of said neural network;receiving the modified pseudo control signal and calculating an inverse response control signal at an inverse response function unit that is based on a model for the process; andtransmitting the inverse response control signal to said actuator device. 15. The method of claim 14 wherein the actuator device comprises a hydraulic actuator. 16. The method of claim 14 wherein the actuator device comprises an electrical actuator. 17. The method of claim 14 wherein the actuator device comprises a plurality of synthetic jets.