Dynamically adaptive fuzzy interval controller

• US 5,758,025 A
• Filed: 06/05/1995
• Issued: 05/26/1998
• Est. Priority Date: 06/05/1995
• Status: Expired due to Fees

• Current Assignee: Wu, Kung Chris
• Original Assignee: Wu, Kung Chris
• Inventors: Wu, Kung Chris
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/461,637
• Time in Patent Office: 1,086 Days
• Field of Search: 318/628, 395/3, 395/10-11, 395/60-61, 395/76, 395/900
• US Class Current: 706/3
• CPC Class Codes: G06N 7/023 Learning or tuning the parameters of a fuzzy system
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

A fuzzy logic controller that is dynamically adaptive to varying operating conditions. The controller remaps the intervals of the membership functions between a normalized universe of discourse and the real axis by use of an input fuzzy gain and an output fuzzy gain. The membership functions are also remapped within the universe of discourse through the use of a sensitivity index. Remapping of the membership functions allows the sensitivity of the controller to be adapted to optimize the controller'"'"'s performance in changing operating conditions. This allows a robust controller to be implemented at low computational expense.

Claims

1. A fuzzy logic controller comprising:

a controlled system,

means for sensing at least one variable from said controlled system,

means for fuzzifying said at least one sensed variable by evaluating a plurality of input membership functions for said at least one sensed variable,

means for converting said at least one fuzzified sensed variable to a plurality of output membership functions,

means for defuzzifying said plurality of output membership functions to generate at least one output variable,

means for outputting said at least one output variable to an actuator, and

means for remapping said plurality of input membership functions within an input universe of discourse and for remapping said output membership functions within an output universe of discourse.

2. The fuzzy logic controller of claim 1 wherein said means for remapping said plurality of input membership functions and for remapping said output membership functions uses a sensitivity index that shifts said plurality of input membership functions and said plurality of output membership functions relative to the centroid of their respective universe of discourse.

3. The fuzzy logic controller of claim 2 wherein said sensitivity index is dynamically calculated during operation of the controller.

4. The fuzzy logic controller of claim 2 wherein a higher value of said sensitivity index causes the controller to be less sensitive to small values of said sensed variable and a lower value of said sensitivity index causes the controller to be more sensitive to small values of said sensed variable.

5. The fuzzy logic controller of claim 2 wherein a higher value of said sensitivity index causes said plurality of input membership functions and said plurality of output membership functions to shift away from the centroid of their respective universe of discourse and a lower value of said sensitivity index causes said plurality of input membership functions and said plurality of output membership functions to shift towards the centroid of their respective universe of discourse.

6. The fuzzy logic controller of claim 2 further comprising a performance optimizer that determines the value of said sensitivity index to optimize the performance of the controller.

7. A fuzzy logic controller comprising:

a controlled system,

means for sensing at least one variable from said controlled system,

means for fuzzifying said at least one sensed variable by evaluating a plurality of input membership functions for said at least one sensed variable,

means for converting said at least one fuzzified sensed variable to a plurality of output membership functions,

means for defuzzifying said plurality of output membership functions to generate at least one output variable,

means for outputting said at least one output variable to an actuator, and

a performance optimizer that modifies at least one of said at least one sensed variable, said at least one output variable, said plurality of input membership functions, and said plurality of output membership functions thereby to optimize the performance of the controller.

8. The fuzzy logic controller of claim 7 wherein said performance optimizer stores the values of said at least one sensed variable for a predetermined period of time and optimizes the performance of the controller based on the stored values of said at least one sensed variable.

9. The fuzzy logic controller of claim 7 wherein said performance optimizer dynamically modifies at least one of said at least one sensed variable, said at least one output variable, said plurality of input membership functions, and said plurality of output membership functions during operation of the controller.

10. The fuzzy logic controller of claim 7 wherein said performance optimizer remaps said plurality of input membership functions within an input universe of discourse and remaps said output membership functions within an output universe of discourse.

11. The fuzzy logic controller of claim 10 wherein said performance optimizer uses a sensitivity index that shifts said plurality of input membership functions and said plurality of output membership functions relative to the centroid of their respective universe of discourse.

12. The fuzzy logic controller of claim 7 wherein said performance optimizer scales said at least one sensed variable by an input fuzzy gain and scales said at least one output variable by an output fuzzy gain.

13. A fuzzy logic controller comprising:

a controlled system,

means for sensing at least one variable from said controlled system,

means for fuzzifying said at least one sensed variable by evaluating a plurality of input membership functions for said at least one sensed variable,

means for converting said at least one fuzzified sensed variable to a plurality of output membership functions,

means for defuzzifying said plurality of output membership functions to generate at least one output variable,

means for outputting said at least one output variable to an actuator,

means for scaling said at least one sensed variable by an input fuzzy gain, and

means for scaling said at least one output variable by an output fuzzy gain wherein said output fuzzy gain is calculated independently from said input fuzzy gain.

14. The fuzzy logic controller of claim 13 wherein said input fuzzy gain and said output fuzzy gain are dynamically calculated during operation of the controller.

15. The fuzzy logic controller of claim 13 wherein said plurality of input membership functions and said plurality of output membership functions are contained in a normalized universe of discourse that ranges from -1 to +1 and wherein said input fuzzy gain and said output fuzzy gain map said at least one sensed variable and said at least one output variable between said normalized universe of discourse and the real axis.

16. The fuzzy logic controller of claim 13 further comprising a performance optimizer that calculates said input fuzzy gain and said output fuzzy gain thereby to optimize the performance of the controller.

17. The fuzzy logic controller of claim 16 wherein said performance optimizer increases the value of the input fuzzy gain or the output fuzzy gain or both the input fuzzy gain and the output fuzzy gain if the magnitude of said at least one sensed variable exceeds a predetermined value.

18. The fuzzy logic controller of claim 13 wherein said output fuzzy gain can be set to zero thereby suppressing any control action from the controller.

19. The fuzzy logic controller of claim 13 wherein said at least one sensed variable comprises a plurality of sensed variables and said at least one output variable comprises a plurality of output variables, said controller further comprising means for scaling said plurality of sensed variables by a plurality of input fuzzy gains and means for scaling said output variables by a plurality of output fuzzy gains wherein each said input fuzzy gain and each said output fuzzy gain are independently calculated.

20. The fuzzy logic controller of claim 19 further comprising a performance optimizer that calculates each of said input fuzzy gains and each of said output fuzzy gains thereby to optimize the performance of the controller.