Probabilistic goal recognition system and method incorporating inferred unobserved actions

• US 7,146,348 B2
• Filed: 11/01/2002
• Issued: 12/05/2006
• Est. Priority Date: 01/22/2002
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Goldman, Robert P., Geib, Christopher W.
• Primary Examiner(s): Hirl, Joseph P.
• Application Number:
  US10/286,398
• Publication Number:
  US 20030139902A1
• Time in Patent Office: 1,495 Days
• Field of Search: 706/45, 706/46, 706/14, 706/61, 715/707
• US Class Current: 706/45
• CPC Class Codes: A61B 5/0002 Remote monitoring of patients using telemetry, e.g. transmission of vital signals via a communication network A61B5/07 takes precedence; transmission systems for measured values G08C; transmission H04B; transmission of digital information H04L; wireless communication networks H04W
• CPC Class Codes: A61B 5/1117 Fall detection
• CPC Class Codes: B01D 61/025 Reverse osmosis; Hyperfiltration
• CPC Class Codes: C02F 1/441 by reverse osmosis
• CPC Class Codes: C02F 2209/40 Liquid flow rate
• CPC Class Codes: C02F 2209/42 Liquid level
• CPC Class Codes: F24F 2006/006 with water treatment
• CPC Class Codes: F24F 6/00 Air-humidification , e.g. cooling by humidification
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06F 21/554 involving event detection and direct action

Abstract

A system and method for recognizing a goal of an actor in an environment by considering hypothesized unobserved actions. The method includes providing a plan library that includes a list of possible goals and corresponding plans. Actions of the actor are observed, and the occurrence of hypothesized unobserved actions are considered. Finally, one or more plans consistent with the observed actions and the consideration of hypothesized unobserved actions are identified.

Claims

1. An automated method of recognizing a goal of an actor in an environment, the method comprising:

providing a plan library including a listing of possible goals and plans to accomplish each of the possible goals, each of the plans comprising a plurality of actions for completing the plan;

observing actions relative to the actor;

considering whether hypothesized unobserved actions have occurred; and

identifying one or more plans consistent with the observed actions and the considered occurrence of hypothesized unobserved actions.

2. The method of claim 1, further comprising:

probabilistically evaluating the identified plans to designate a most likely goal of the actor.

3. The method of claim 2, wherein the step of probabilistically evaluating includes determining a probability distribution over the set of consistent plans.

4. The method of claim 1, wherein considering whether hypothesized unobserved actions have occurred includes:

identifying at least one primitive action in one of the plans that is not otherwise one of the observed actions;

designating the at least one primitive action as a potential hypothesized unobserved action; and

evaluating whether the potential hypothesized unobserved action is likely to have occurred.

5. The method of claim 4, wherein evaluating the potential hypothesized unobserved action includes:

probabilistically determining whether the potential hypothesized unobserved action is enabled by one or more of the observed actions.

6. The method of claim 4, wherein evaluating the potential at least one hypothesized unobserved action includes:

probabilistically determining whether the potential hypothesized unobserved action is consistent with an observed state change.

7. The method of claim 4, wherein evaluating the potential hypothesized unobserved action includes:

establishing a probabilistic threshold value for unobserved actions;

probabilistically determining a likelihood of the potential hypothesized unobserved action having occurred relative to the observed actions; and

determining whether the determined likelihood is less than the probabilistic threshold value.

8. The method of claim 4, wherein a plurality of hypothesized unobserved actions are considered.

9. The method of claim 1, wherein the plans in the plan library are modeled as partially ordered plan graphs.

10. The method of claim 1, wherein the actor is a human actor operating in a daily living environment.

11. The method of claim 10, wherein observing actions includes:

providing a plurality of sensors in the daily living environment, each adapted to provide information indicative of at least one of an action of the actor and an event in the environment; and

reviewing information from the plurality of sensors.

12. The method of claim 11, wherein the plurality of sensors includes at least one sensor selected from the group consisting of a motion detector, pressure pad, door latch sensor, panic button, toilet-flush sensor, microphone, camera, fall sensor, heart rate sensor, blood pressure monitor sensor, glucose monitor sensor, moisture sensor, and appliance actuator sensor.

13. The method of claim 1, wherein the actor is an unauthorized user operating in a computer network environment.

14. The method of claim 13, wherein observing actions includes:

monitoring usage of the computer network.

15. A computer assisted system for recognizing a goal of an actor in an environment, the system comprising:

an information input device for generating information indicative of actions of the actor;

a plan library including a listing of possible goals, wherein each goal includes a sub-listing of at least one plan for accomplishing the corresponding goal upon completing the corresponding plan, each plan including a sub-listing of primitive actions for performing that plan; and

a controller adapted to;

receive information from the information input device,

record observed actions,

consider whether hypothesized unobserved actions have occurred,

identify one or more plans consistent with the observed actions and the considered occurrence of hypothesized unobserved actions.

16. The system of claim 15, wherein the controller is further adapted to probabilistically evaluate the identified plans for designating a most likely goal of the actor.

17. The system of claim 16, wherein the controller is further adapted to determine a probabilistic distribution over the set of consistent plans.

18. The system of claim 15, wherein the controller is further adapted to:

identify at least one primitive action in one of the plans that is not otherwise one of the recorded observed actions;

designate the at least one primitive action as a potential hypothesized unobserved action; and

evaluate whether the potential hypothesized unobserved action is likely to have occurred.

19. The system of claim 18, wherein the controller is further adapted to:

probabilistically determine whether the potential hypothesized unobserved action is enabled by one or more of the observed actions.

20. The system of claim 18, wherein the controller is further adapted to:

probabilistically determine whether the potential hypothesized unobserved action is consistent with an observed state change.

21. The system of claim 18, wherein the controller is further adapted to:

store a probabilistic threshold value for unobserved actions;

probabilistically determine a likelihood of the potential hypothesized unobserved action having occurred relative to the observed actions; and

determine whether the determined likelihood is less than the probabilistic threshold value.

22. The system of claim 15, wherein the plans in the plan library are modeled as partially ordered plan graphs.

23. The system of claim 15, wherein the controller is further adapted to perform monitoring functions relating to a human actor operating in a living space environment.

24. The system of claim 23, wherein the information input device includes a plurality of sensors adapted to sense activities of the human actor in the living space.

25. The system of claim 24, further comprising:

an effector linked to the controller and adapted to control operation of a device in the living space.

26. The system of claim 15, wherein the controller is further adapted to perform a response planning function based upon a determined goal of the actor.

27. The system of claim 15, wherein controller is further adapted to perform monitoring functions of an unauthorized user operating in a computer network environment.