Apparatus and method for optimizing the performance of computer tasks using multiple intelligent agents having varied degrees of domain knowledge

• US 6,192,354 B1
• Filed: 03/21/1997
• Issued: 02/20/2001
• Est. Priority Date: 03/21/1997
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Cragun, Brian John, Bigus, Joseph Phillip, Delp, Helen Roxlo
• Primary Examiner(s): Hafiz, Tariq R.
• Assistant Examiner(s): Davis, George
• Application Number:
  US08/826,107
• Time in Patent Office: 1,432 Days
• Field of Search: 706/10, 706/45, 706/46, 706/16, 705/26, 705/35, 705/37, 395/200.32
• US Class Current: 706/46
• CPC Class Codes: G06N 5/043 Distributed expert systems; Blackboards

Abstract

The performance of a given computer task is optimized by utilizing a plurality of intelligent agents suited to perform the computer task but having varied degrees of domain knowledge. Based upon an objective criteria that may be determined for a given situation, one of the plurality of intelligent agents may be selected and dispatched to perform the task, thereby optimizing the performance of the computer task for a wide variety of situations.

Claims

1. A method of handling a computer task using an intelligent agent, the method comprising the steps of:

(a) based upon an objective criteria, selecting for dispatch a selected intelligent agent from a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, wherein a non-selected intelligent agent is retained after selection of the selected intelligent agent; and

(b) dispatching the selected intelligent agent to handle the computer task.

2. The method of claim 1, wherein the selecting step includes the step of adaptively selecting the selected intelligent agent using a reinforcement learning algorithm.

3. The method of claim 2, further comprising the steps of:

(a) obtaining performance information relating to the performance of the selected intelligent agent in handling the computer task; and

(b) supplying the performance information to the reinforcement learning algorithm.

4. The method of claim 2, wherein the reinforcement learning algorithm is implemented in an adaptive heuristic critic neural network.

5. The method of claim 1, wherein the selecting step includes the steps of:

(a) matching each of the plurality of intelligent agents with a value of the objective criteria;

(b) determining a selected value of the objective criteria; and

(c) selecting as the selected intelligent agent an intelligent agent matching the selected value of the objective criteria.

6. The method of claim 5, wherein the selecting step further includes the step of retrieving information for a selected computer task, wherein the determining step determines the selected value of the objective criteria using the retrieved information.

7. The method of claim 1, wherein each of the plurality of intelligent agents is configured to conduct negotiations in an electronic commerce application, and wherein the domain knowledge for each of the plurality of intelligent agents is related to the autonomy delegated thereto.

8. The method of claim 7, wherein the plurality of intelligent agents includes a semi-autonomous agent, a fully-autonomous agent, and a fully-dependent agent.

9. The method of claim 7, wherein the objective criteria includes a risk that a dispatched agent is subjected to in negotiations.

10. An apparatus for handling a computer task, comprising:

a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, wherein, based upon an objective criteria, a selected agent from the plurality of intelligent agents is selected and dispatched to handle the computer task, and wherein a non-selected intelligent agent is retained after selection of the selected intelligent agent.

11. The apparatus of claim 10, further comprising an agent manager configured to dispatch the selected intelligent agent based upon the objective criteria.

12. The apparatus of claim 11, further comprising a reinforcement learning module, coupled to the agent manager and configured to adaptively select intelligent agents based upon the performance of the plurality of intelligent agents in handling the computer task.

13. The apparatus of claim 12, wherein the reinforcement learning module comprises an adaptive heuristic critic neural network.

14. The apparatus of claim 11, wherein the agent manager is configured to retrieve information for a selected computer task, determine a selected value for the objective criteria for the selected computer task, and select as the selected intelligent agent one of the plurality of intelligent agents which is matched with the selected value of the objective criteria.

15. The apparatus of claim 10, wherein each of the plurality of intelligent agents is configured to conduct negotiations in an electronic commerce application, and wherein the domain knowledge for each of the plurality of intelligent agents is related to the autonomy delegated thereto.

16. The apparatus of claim 15, wherein the plurality of intelligent agents includes a semi-autonomous agent, a fully-autonomous agent, and a fully-dependent agent.

17. The apparatus of claim 15, wherein the objective criteria includes a risk that a dispatched agent is subjected to in negotiations.

18. A program product comprising:

(a) a program configured to handle a computer task using an intelligent agent, the program comprising a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, wherein, based upon an objective criteria, a selected agent from the plurality of intelligent agents is selected and dispatched to handle the computer task, and wherein a non-selected intelligent agent is retained after selection of the selected intelligent agent; and

(b) a signal bearing media bearing the program.

19. The program product of claim 18, wherein the signal bearing media is transmission type media.

20. The program product of claim 18, wherein the signal bearing media is recordable media.

21. A method of handling a computer task on a remote computer system using an intelligent agent, the method comprising the steps of:

(a) determining a risk for the remote computer system;

(b) based upon the risk for the remote computer system, selecting a selected intelligent agent from a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task on the remote computer system; and

(c) dispatching the selected intelligent agent to handle the computer task.

22. The method of claim 21, further comprising the step of matching each of the plurality of intelligent agents with a risk level.

23. The method of claim 22, wherein the matching step includes the step of adaptively matching each intelligent agent based upon the actual performance of the plurality of intelligent agents.

24. A method of handling a computer task using an intelligent agent, the method comprising the steps of:

(a) based upon an objective criteria, adaptively selecting a selected intelligent agent from a plurality of intelligent agents having varied degrees of domain knowledge using a reinforcement learning algorithm implemented in an adaptive heuristic critic neural network, wherein the plurality of intelligent agents are configured to handle a common computer task; and

(b) dispatching the selected intelligent agent to handle the computer task.

25. A method of handling a computer task using an intelligent agent, the method comprising the steps of:

(a) based upon an objective criteria, selecting a selected intelligent agent from a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, wherein each of the plurality of intelligent agents is configured to conduct negotiations in an electronic commerce application, and wherein the domain knowledge for each of the plurality of intelligent agents is related to the autonomy delegated thereto; and

(b) dispatching the selected intelligent agent to handle the computer task.

26. The method of claim 25, wherein the plurality of intelligent agents includes a semi-autonomous agent, a fully-autonomous agent, and a fully-dependent agent.

27. The method of claim 25, wherein the objective criteria includes a risk that a dispatched agent is subjected to in negotiations.

28. An apparatus for handling a computer task, comprising:

a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task; and

a reinforcement learning module configured to adaptively select, based upon an objective criteria, a selected intelligent agent from the plurality of intelligent agents to dispatch to handle the computer task, the reinforcement learning module including an adaptive heuristic critic neural network.

29. The apparatus of claim 28, further comprising an agent manager, coupled to the reinforcement learning module and configured to dispatch the selected intelligent agent.

30. An apparatus for handling a computer task, comprising:

a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, wherein, based upon an objective criteria, a selected agent from the plurality of intelligent agents is dispatched to handle the computer task, wherein each of the plurality of intelligent agents is configured to conduct negotiations in an electronic commerce application, and wherein the domain knowledge for each of the plurality of intelligent agents is related to the autonomy delegated thereto.

31. The apparatus of claim 30, wherein the plurality of intelligent agents includes a semi-autonomous agent, a fully-autonomous agent, and a fully-dependent agent.

32. The apparatus of claim 30, wherein the objective criteria includes a risk that a dispatched agent is subjected to in negotiations.

33. An apparatus for handling a computer task on a remote computer system, comprising:

a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task on a remote computer system, and wherein, based upon a risk determined for the remote computer system, a selected agent from the plurality of intelligent agents is dispatched to handle the computer task.

34. A program product comprising:

(a) a program configured to handle a computer task using an intelligent agent, the program comprising a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, and a reinforcement learning module configured to adaptively select, based upon an objective criteria, a selected intelligent agent from the plurality of intelligent agents to dispatch to handle the computer task, the reinforcement learning module including an adaptive heuristic critic neural network; and

(b) a signal bearing media bearing the program.

35. A program product comprising:

(a) a program configured to handle a computer task using an intelligent agent, the program comprising a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task, wherein, based upon an objective criteria, a selected agent from the plurality of intelligent agents is dispatched to handle the computer task, wherein each of the plurality of intelligent agents is configured to conduct negotiations in an electronic commerce application, and wherein the domain knowledge for each of the plurality of intelligent agents is related to the autonomy delegated thereto; and

(b) a signal bearing media bearing the program.

36. A program product comprising:

(a) a program configured to handle a computer task on a remote computer system using an intelligent agent, the program comprising a plurality of intelligent agents having varied degrees of domain knowledge, wherein the plurality of intelligent agents are configured to handle a common computer task on a remote computer system, and wherein, based upon a risk determined for the remote computer system, a selected agent from the plurality of intelligent agents is dispatched to handle the computer task; and

(b) a signal bearing media bearing the program.