×

Method for rolling back an expert system

  • US 5,170,464 A
  • Filed: 11/15/1991
  • Issued: 12/08/1992
  • Est. Priority Date: 01/26/1990
  • Status: Expired due to Fees
First Claim
Patent Images

1. A method for rolling back the state of a rule-based expert system on an inference engine, comprising the steps of:

  • providing an inference stack of state records;

    providing an execution stack of execution state records, each of which identifies a state for the inference engine;

    responsive to user selection of at least a first breakpoint condition definition, establishing at least a first breakpoint for the inference engine;

    operating the inference engine to generate a chain through a rule/variable network by removing inference state records and execution state records from the inference stack and the execution stack, respectively;

    advancing an execution state number with each execution state record removed from the execution stack;

    for each inference state record removed from the inference stack, copying the inference state record with a current execution state number to an inference history queue;

    for each execution state record removed from the execution stack, copying the execution state record with a current execution state number to an execution history queue;

    upon the inference engine encountering a breakpoint during operation, adding an entry to a breakpoint history queue;

    accepting a breakpoint condition definition for rollback of the execution state from a user;

    comparing the accepted breakpoint condition definition and the entries in the breakpoint history queue to determine a desired execution state and its state number;

    resetting the inference and execution stacks by moving state records and copied execution state records from the inference history queue and execution history queue to the inference stack and execution stack, respectively;

    for each copied execution state record replaced from the execution history queue to the execution stack, reducing the execution state number and retracing the chain through the rule/variable network; and

    discontinuing resetting of the inference and execution stacks when the execution state number equals the state number for the desired execution state.

View all claims
  • 0 Assignments
Timeline View
Assignment View
    ×
    ×