Method for operating a real-time expert system in a graphical programming environment

• US 5,133,045 A
• Filed: 11/06/1991
• Issued: 07/21/1992
• Est. Priority Date: 07/13/1990
• Status: Expired due to Term

• Current Assignee: The MathWorks, Inc.
• Original Assignee: Integrated Systems Incorporated
• Inventors: Gaither, Shawn A., Shah, Sunil C.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/790,127
• Time in Patent Office: 258 Days
• Field of Search: 395/51, 395/75, 395/76, 395/68, 395/700, 395/12
• US Class Current: 706/46
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A method is disclosed for operating a real-time multirate and discrete even computer system for simulation and automatic code generation of modelled systems requiring both expert system technology and conventional algorithms. Simulation of a system to be modelled proceeds on data inputs to yield outputs as responses of the simulated system to data inputs for assembling an organized catalog of blocks, both conventional and expert system, based on the operational rates and time skews. These blocks can be assembled for automatic code generation so that all blocks operating with identical rates and time skews can be organized within a single subsystem. Code is then generated for each subsystem, including code for a scheduler to run the entire system thus modelled. Each expert system block generates its own inference engine, and set of parameters representing the knowledge base, and storage allocation for the data base.

Claims

1. A method of operating a multirate real-time expert system, comprising:

providing a computer having at least one central processing unit and memory for storing data and software;

storing, in said memory, system definition data representing said multirate real-time system as a set of interconnected functional blocks, including data specifying computations to be performed each time one of said functional blocks is executed, said system definition data also specifying external input signals and external output signals for said multirate real-time system;

defining subsets of said functional blocks as subsystems of the multirate real-time system;

at least one subset of said functional blocks comprising a rule-based expert subsystem, said data specifying computations to be performed by said rule-based expert subsystem including data representing a plurality of logical rules, rule selection software for selecting a subset of said rules for execution, and rule execution software for executing the selected subset of rules to produce a set of output signal values;

defining in said memory a double buffer interface for storing input and output signal values associated with said functional blocks, including defining;

(A) an output signal buffer for holding output signal values generated by said functional blocks during execution thereof, and (B) a sample and hold buffer for each said subsystem;

scheduling and executing said specified computations associated with said defined subsystems; and

prior to executing said specified computations associated with each said defined subsystem, copying into said sample and hold buffer for each said defined subsystem all input signal values to be used by said specified computations, and holding said input signal values copied into said sample and hold buffer invariant during execution of said specified computations associated with each said defined subsystem;

whereby a consistent set of input signal values is presented to said rule-based expert system regardless of changing values of associated ones of said input signals.

2. The method of claim 1,

wherein said output signal buffer comprises:

two equal size output buffers for each defined subsystem, each equal size output buffer holding output signal values and internal state variables of said defined subsystem; and

flag means, for each defined subsystem, having a value indicating which of said two equal size output buffers is used during each execution of said functional blocks in said subsystem;

said scheduling step including changing the value of said flag means associated with each defined subsystem prior to execution of said functional blocks in said defined subsystem, and copying into said sample and hold buffer specified output signals from the one of said two equal size output buffers of other defined subsystems not currently designated by said flag means.

3. The method of claim 1,

defining for each said subsystem a computation initiation condition comprising a computational trigger condition for asynchronous subsystems and a computational repetition rate and skew period for synchronous subsystems;

said rule-based expert subsystem comprising a synchronous subsystem having an associated computational repetition rate and skew period;

said scheduling and executing step including;

defining said computational repetition rate and skew period for said rule-based expert subsystem based on a specified output signal schedule of times at which output signals from said rule-based expert subsystem are needed; and

scheduling initiation of execution of said specified computations associated with said rule-based expert subsystem based on said defined output signal schedule.

4. The method of claim 1,

defining for each of a plurality of said subsystems a computational repetition rate and skew period based on a specified output signal schedule of times at which output signals from subsystems are needed;

and

scheduling initiation of execution of said specified computations associated with said plurality of said subsystems based on said defined output signal schedule.

5. The method of claim 1, said data representing a plurality of logical rules includes a set of pre-computed interrelations between data and rules which is stored as a linked list of interrelations used by said rule selection software for selecting a subset of said rules for execution.

6. The method of claim 1, said rule selection software includes software for forming a selectable hierarchical ordering of heuristics by which rules are selected.

7. The method of claim 1, said rule selection software allows confidence levels and priority levels associated with said plurality of logical rules to vary with time.

8. The method of claim 1, said multirate real-time system including a plurality of rule-based expert subsystems.

9. The method of claim 7, wherein each of said plurality of rule-based expert subsystems comprises a distinct synchronous subsystem of said multirate real-time system having an associated computational repetition rate and skew period.

10. A method of simulating operation of a multirate real-time expert system, comprising:

providing a computer having at least one central processing unit and memory for storing data and software;

storing, in said memory, system definition data representing said multirate real-time system as a set of interconnected functional blocks, including data specifying computations to be performed each time one of said functional blocks is executed, said system definition data also specifying external input signals and external output signals for said multirate real-time system;

said specified computations simulating operation of corresponding portions of said multirate real-time expert system;

defining subsets of said functional blocks as subsystems of the multirate real-time system;

at least one subset of said functional blocks comprising a rule-based expert subsystem, said data specifying computations to be performed by said rule-based expert subsystem including data representing a plurality of logical rules, rule selection software for selecting a subset of said rules for execution, and rule execution software for executing the selected subset of rules to produce a set of output signal values;

defining in said memory a double buffer interface for storing input and output signal values associated with said functional blocks, including defining;

(A) an output signal buffer for holding output signal values generated by said functional blocks during execution thereof, and (B) a sample and hold buffer for each said subsystem;

scheduling and executing said specified computations associated with said defined subsystems; and

prior to executing said specified computations associated with each said defined subsystem, copying into said sample and hold buffer for each said defined subsystem all input signal values to be used by said specified computations, and holding said input signal values copied into said sample and hold buffer invariant during execution of said specified computations associated with each said defined subsystem;

whereby a consistent set of input signal values is presented to said rule-based expert system regardless of changing values of associated ones of said input signals.

11. The method of claim 10,

wherein said output signal buffer comprises:

two equal size output buffers for each defined subsystem, each equal size output buffer holding output signal values and internal state variables of said defined subsystem; and

flag means, for each defined subsystem, having a value indicating which of said two equal size output buffers is used during each execution of said functional blocks in said subsystem;

said scheduling step including changing the value of said flag means associated with each defined subsystem prior to execution of said functional blocks in said defined subsystem, and copying into said sample and hold buffer specified output signals from the one of said two equal size output buffers of other defined subsystems not currently designated by said flag means.

12. The method of claim 10,

defining for each said subsystem a computation initiation condition comprising a computational trigger condition for asynchronous subsystems and a computational repetition rate and skew period for synchronous subsystems;

said rule-based expert subsystem comprising a synchronous subsystem having an associated computational repetition rate and skew period;

said scheduling and executing step including;

defining said computational repetition rate and skew period for said rule-based expert subsystem based on a specified output signal schedule of times at which output signals from said rule-based expert subsystem are needed; and

scheduling initiation of execution of said specified computations associated with said rule-based expert subsystem based on said defined output signal schedule.

13. The method of claim 10,

defining for each of a plurality of said subsystems a computational repetition rate and skew period based on a specified output signal schedule of times at which output signals from subsystems are needed;

and

scheduling initiation of execution of said specified computations associated with said plurality of said subsystems based on said defined output signal schedule.

14. The method of claim 10, said data representing a plurality of logical rules including a set of pre-computed interrelations between data and rules which is stored as a linked list of interrelations used by said rule selection software for selecting a subset of said rules for execution.

15. The method of claim 10, said rule selection software including software for forming a selectable hierarchical ordering of heuristics by which rules are selected.

16. The method of claim 10, wherein said rule selection software allows confidence levels and priority levels associated with said plurality of logical rules to vary with time.

17. The method of claim 10, said multirate real-time system including a plurality of rule-based expert subsystems.

18. The method of claim 17, wherein each of said plurality of rule-based expert subsystems comprises a distinct synchronous subsystem of said multirate real-time system having an associated computational repetition rate and skew period.

19. A multirate real-time expert computer system, comprising:

a computer having at least one central processing unit and memory for storing data and software;

system definition means coupled to said memory for storing, in said memory, system definition data representing said multirate real-time system as a set of interconnected functional blocks, including data specifying computations to be performed each time one of said functional blocks is executed, said system definition data also specifying external input signals and external output signals for said multirate real-time system;

said system definition means defining subsets of said functional blocks as subsystems of the multirate real-time system;

at least one subset of said functional blocks comprising a rule-based expert subsystem, said data specifying computations to be performed by said rule-based expert subsystem including data representing a plurality of logical rules, rule selection software for selecting a subset of said rules for execution, and rule execution software for executing the selected subset of rules to produce a set of output signal values;

a double buffer interface in said memory for storing input and output signal values associated with said functional blocks, including (A) an output signal buffer for holding output signal values generated by said functional blocks during execution thereof, and (B) a sample and hold buffer for each said subsystem;

an execution scheduler which schedules and initiates execution of said specified computations associated with said defined subsystems; and

sample and hold means, responsive to said execution scheduler, for copying into said sample and hold buffer, prior to executing said specified computations associated with each said defined subsystem, all input signal values to be used by said specified computations, and for holding said input signal values copied into said sample and hold buffer invariant during execution of said specified computations associated with each said defined subsystem;

whereby a consistent set of input signal values is presented to said rule-based expert system regardless of changing values of associated ones of said input signals.

20. The multirate real-time expert computer system of claim 19,

wherein said output signal buffer comprises:

two equal size output buffers for each defined subsystem, each equal size output buffer holding output signal values and internal state variables of said defined subsystem; and

flag means, for each defined subsystem, having a value indicating which of said two equal size output buffers is used during each execution of said functional blocks in said subsystem;

said execution scheduler changing the value of said flag means associated with each defined subsystem prior to execution of said functional blocks in said defined subsystem, and copying into said sample and hold buffer specified output signals from the one of said two equal size output buffers of other defined subsystems not currently designated by said flag means.

21. The multirate real-time expert computer system of claim 19,

said system definition means defining for each said subsystem a computation initiation condition comprising a computational trigger condition for asynchronous subsystems and a computational repetition rate and skew period for synchronous subsystems;

said rule-based expert subsystem comprising a synchronous subsystem having an associated computational repetition rate and skew period;

said execution scheduler (A) defining said computational repetition rate and skew period for said rule-based expert subsystem based on a specified output signal schedule of times at which output signals from said rule-based expert subsystem are needed, and (B) scheduling initiation of execution of said specified computations associated with said rule-based expert subsystem based on said defined output signal schedule.

22. The multirate real-time expert computer system of claim 19,

said system definition means defining for each of a plurality of subsystems a computational repetition rate and skew period based on a specified output signal schedule of times at which output signals from subsystems are needed;

and

said execution scheduler scheduling initiation of execution of said specified computations associated with said plurality of said subsystems based on said defined output signal schedule.

23. The multirate real-time expert computer system of claim 19, said data representing a plurality of logical rules including a set of pre-computed interrelations between data and rules which is stored as a linked list of interrelations used by said rule selection software for selecting a subset of said rules for execution.

24. The multirate real-time expert computer system of claim 19, said rule selection software including software for forming a selectable hierarchical ordering of heuristics by which rules are selected.

25. The multirate real-time expert computer system of claim 19, wherein said rule selection software allows confidence levels and priority levels associated with said plurality of logical rules to vary with time.

26. The multirate real-time expert computer system of claim 19, said multirate real-time system including a plurality of rule-based expert subsystems.

27. The multirate real-time expert computer system of claim 26, wherein each of said plurality of rule-based expert subsystems comprises a distinct synchronous subsystem of said multirate real-time system having an associated computational repetition rate and skew period.