Methods and systems for estimating subject cost from surveillance
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06N-005/00
G06F-001/00
출원번호
US-0405698
(2012-02-27)
등록번호
US-8959042
(2015-02-17)
발명자
/ 주소
Singliar, Tomas
Margineantu, Dragos Dorin
출원인 / 주소
The Boeing Company
대리인 / 주소
Armstrong Teasdale LLP
인용정보
피인용 횟수 :
0인용 특허 :
5
초록▼
A computer-based method for analyzing the costs of agent's behaviors is described. The method includes storing data relating to a previously observed behavior of at least one of an agent of interest and at least one agent that can be assumed to hold similar utilities to the agent of interest, such t
A computer-based method for analyzing the costs of agent's behaviors is described. The method includes storing data relating to a previously observed behavior of at least one of an agent of interest and at least one agent that can be assumed to hold similar utilities to the agent of interest, such that an agent class is defined, deriving with a processing device and based on the stored data, a resolving utility function, and observing a sequence of behavior of the agent of interest. The method also includes inputting the observed behavior sequence to an analyzer, deriving with a processing device and based on the observed sequence of behavior, a set of costs that the agent of interest incurred for their observed behavior, and comparing the resolving utility function derived from stored data to the set of costs derived from the observed sequence of behavior to determine anomalous behavior.
대표청구항▼
1. A method for analyzing the costs of an agent, said method comprising: storing, in a computer memory, data relating to a previously observed behavior of at least one of an agent of interest and at least one agent that can be assumed to hold similar utilities to the agent of interest, such that an
1. A method for analyzing the costs of an agent, said method comprising: storing, in a computer memory, data relating to a previously observed behavior of at least one of an agent of interest and at least one agent that can be assumed to hold similar utilities to the agent of interest, such that an agent class is defined;deriving with a processing device and based on the stored data, a resolving utility function;observing a sequence of behavior of the agent of interest;inputting the observed behavior sequence to an analyzer;deriving with a processing device and based on the observed sequence of behavior, a set of costs that the agent of interest incurred for their observed behavior;determining at least one percept of the agent regarding the set of costs; andcomparing the resolving utility function derived from stored data to the set of costs derived from the observed sequence of behavior and the at least one percept regarding the set of costs to determine anomalous behavior. 2. The method according to claim 1 further comprising: deriving specific goals of the agent for the observed sequence of behavior; andpredicting future actions of the agent of interest by determining a minimum-cost action sequence which leads the agent of interest to achieve its stated or derived goals from its present state. 3. The method according to claim 2 wherein the step of predicting future actions of the agent of interest comprises analyzing future actions of the agent of interest based on the minimum-cost action sequence for the agent of interest. 4. The method according to claim 2 wherein predicting future actions of the agent of interest comprises assigning a probability of being a goal of the agent of interest to each member of the agent class. 5. The method according to claim 2 wherein inputting the observed behavior sequence to an analyzer comprises inputting the observed behavior sequence into a processing device executing an executable algorithm. 6. The method according to claim 2 further comprising determining if an agent of interest does not conform to the predicted future actions. 7. The method according to claim 1 wherein deriving a resolving utility function further comprises deriving a cost for each possible action of the agent of interest in a mesh space. 8. The method according to claim 7 wherein deriving a cost for each possible action of the agent of interest comprises: associating a cost function with each state and action pair defining a possible action;calculating a cost-to-go function expressing an unknown shortest distance associated with each destination; anddenoting the cost incurred by the unique action of moving to each destination. 9. The method according to claim 8 further comprising: adding a path cost constraint for each observed path, wherein the constraint is specifying that the total cost for a path be the sum of action and state costs for all action and states on the path, the path being an interleaved sequence of actions and states representing the agent's action sequence;inserting a “Bellman” constraint for each state on the path and for each neighboring state on the path; andsolving the cost functions using a maximization objective that is the sum of terms expressing the margin for each action taken. 10. The method according to claim 8 wherein a cost is an expected cost that extends formulations to a probabilistic case. 11. The method according to claim 1 further comprising at least one of: disposing of the action example as uninteresting such that it will not be further analyzed;saving the action example and explaining the example away by other costs afforded to the agent of interest; andconstructing a basis function that explains the behavior of the agent of interest. 12. One or more computer-readable storage media having computer-executable instructions embodied thereon, wherein when executed by at least one processor, the computer-executable instructions cause the at least one processor to: store, in a computer memory, data relating to previously observed behavior of at least one of an agent of interest and agents that can be assumed to hold similar utilities to the agent of interest, thereby defining an agent class;derive with a processing device and based on the stored data a resolving utility function;observe a sequence of behavior of the agent of interest;input the observed behavior sequence to an analyzer;derive with a processing device and based on the observed sequence of behavior, a resolving utility function;derive with a processing device and based on the observed sequence of behavior, a set of costs that the agent of interest incurred for their observed behavior;determine at least one percept of the agent regarding the set of costs; andcompare the resolving utility function derived from stored data to the set of costs derived from the observed sequence of behavior and the at least one percept regarding the set of costs to determine anomalous behavior. 13. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 12 wherein when executed by at least one processor, the computer-executable instructions cause the at least one processor to: derive specific goals of the agent for the observed sequence of behavior; andpredict future actions of the agent of interest based on a minimum set of costs for the agent of interest and the derived goals. 14. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 13 wherein to predict future actions of the agent of interest, the computer-executable instructions cause the at least one processor to analyze future actions of the agent of interest based on the minimum costs for the agent of interest. 15. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 13 wherein to predict future actions of the agent of interest, the computer-executable instructions cause the at least one processor to assign a probability of being a goal of the agent of interest to each member of a set of possible goals. 16. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 12 wherein the computer-executable instructions cause the at least one processor to point out an agent of interest that does not conform to the predicted future actions. 17. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 12 wherein to derive a resolving utility function, the computer-executable instructions cause the at least one processor to derive a cost for each possible action of the agent of interest in a mesh space. 18. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 12 wherein the computer-executable instructions cause the at least one processor to: associate a cost function with each state and action pair defining a possible action;calculate a cost-to-go function expressing an unknown shortest distance associated with each destination; anduse a function to denote the cost incurred by the unique action of moving to each destination. 19. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 18 wherein the computer-executable instructions cause the at least one processor to: add a constraint for each observed path, the constraint specified as an interleaving sequence of states and a cost;insert a “Bellman” constraint for each state on the path, and each neighboring state on the path; andsolve the cost functions using a maximization objective that is the sum of terms expressing the margin for each action taken. 20. One or more computer-readable storage media having computer-executable instructions embodied thereon according to claim 12 wherein the computer-executable instructions cause the at least one processor to at least one of: dispose of the action example as uninteresting such that it will not be further analyzed;save the action example and explaining the example away by other costs afforded to the agent of interest and not extending a basis function set; andconstruct a basis function that explains the behavior of the agent of interest and extending the basis function set. 21. A method comprising: utilizing the output of a tracking system to track and store observations of behaviors of an agent;estimating an optimal path for the agent based on the stored observations of agent behavior;deriving with a processing device and based on the stored data, a resolving utility function;determining at least one percept of the agent regarding a set of costs that the agent incurred for the observed behavior; anddetermining anomalous behavior of the agent based at least on the set of costs, the at least one percept, and the derived resolving utility function.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (5)
Aviv David G., Abnormality detection and surveillance system.
Wyschogrod Daniel (Brookline MA) Wood Loren (Lexington MA) Sturdy James L. (Leominster MA) Schultz Hayden B. (Maynard MA) Sasiela Richard J. (Sudbury MA) Marquis Douglas V. (Framingham MA) Harman ; I, Airport surface surveillance system.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.