Using long-range dynamics and mental-state models to assess collision risk for early warning

• US 7,792,641 B2
• Filed: 06/12/2007
• Issued: 09/07/2010
• Est. Priority Date: 06/12/2007
• Status: Active Grant

• Current Assignee: Uatc, LLC (Uber Technologies, Inc.)
• Original Assignee: Palo Alto Research Center, Inc. (Xerox Holdings Corp.)
• Inventors: Liu, Juan, Reich, James E., Greene, Daniel H.
• Primary Examiner(s): NGUYEN, TAN QUANG
• Application Number:
  US11/818,122
• Publication Number:
  US 20080312830A1
• Time in Patent Office: 1,183 Days
• Field of Search: 701/301.1, 701/300, 340/436, 340/903
• US Class Current: 701/301
• CPC Class Codes: G08G 1/165 for passive traffic, e.g. including static obstacles, trees
• CPC Class Codes: G08G 1/166 for active traffic, e.g. moving vehicles, pedestrians, bikes

Abstract

One embodiment of the present invention provides a system that for facilitating assessment of collision between a primary principal and a non-primary principal for early warning. During operation, the system periodically performs the following operations: The system obtains a current observation of the primary principal and non-primary principal. The system then assesses one or more future states for the primary and non-primary principals, respectively, based on: the current observation of the primary and non-primary principals, a dynamics model of the primary principal, and a mental-state model of a person associated with the primary principal. The system further produces one or more results which indicate an assessment of collision between the primary and non-primary principals.

Claims

1. A method for facilitating assessment of collision between a primary principal and a non-primary principal for early warning, the method comprising periodically performing:

obtaining a current observation of the primary principal and non-primary principal;

assessing one or more future states for the primary and non-primary principals, respectively, based on;

the current observation of the primary and non-primary principals,

a dynamics model of the primary principal, and

a mental-state model of a person associated with the primary principal; and

producing one or more results which indicate an assessment of collision between the primary and non-primary principals.

2. The method of claim 1,

wherein assessing the future states for the primary and non-primary principals comprises performing sequential Bayesian filtering based on the current observation and past observations of the primary and non-primary principals, respectively.

3. The method of claim 2,

wherein performing sequential Bayesian filtering comprises performing particle filtering.

4. The method of claim 2,

wherein performing sequential Bayesian filtering comprises performing Interacting Multiple Model (IMM) filtering.

5. The method of claim 1,

wherein the dynamics model of the primary principal describes the movements of the primary principal based on a scenario.

6. The method of claim 1,

wherein assessing the future states of the non-primary principal is based on a dynamics model which describes the movements of the non-primary principal based on a scenario.

7. The method of claim 1,

wherein the mental-state model includes an “

alert”

state and a “

not-alert”

state; and

wherein the mental-state model specifies a first probability of transition from the “

alert”

state to the “

not-alert”

state and a second probability of transition from the “

not-alert”

state to the “

alert”

state.

8. The method of claim 1,

wherein the mental-state model includes a “

rational-decision”

state and an “

irrational-decision”

state; and

wherein the mental-state model specifies a first probability of transition from the “

rational-decision”

state to the “

irrational-decision”

state and a second probability of transition from the “

irrational-decision”

state to the “

rational-decision”

state.

9. The method of claim 1,

wherein a state of the primary or non-primary principal includes one or more of:

a position;

a velocity; and

a mental state of the person associated with the primary or non-primary principal.

10. The method of claim 1,

where the results include one or more of:

a probability of collision,

a predicted time of collision,

a predicted location of collision,

a predicted benefit of collision warning, and

an estimated prediction accuracy.

11. A system for facilitating assessment of collision between a primary principal and a non-primary principal for early warning, the system comprising:

a specialized assessment mechanism, comprising;

a data obtaining mechanism configured to obtain a current observation of the primary principal and non-primary principal;

a computation mechanism configured to assess one or more future states for the primary and non-primary principals, respectively, based on;

the current observation of the primary and non-primary principals,

a dynamics model of the primary principal, and

a mental-state model of a person associated with the primary principal; and

a result producing mechanism configured to produce one or more results which indicate an assessment of collision between the primary and non-primary principals.

12. The system of claim 11,

wherein while assessing the future states for the primary and non-primary principals, the computation mechanism is configured to perform sequential Bayesian filtering based on the current observation and past observations of the primary and non-primary principals, respectively.

13. The system of claim 12,

wherein while performing sequential Bayesian filtering, the computation mechanism is configured to perform particle filtering.

14. The system of claim 12,

wherein while performing sequential Bayesian filtering, the computation mechanism is configured to perform Interacting Multiple Model (IMM) filtering.

15. The system of claim 11,

wherein the dynamics model of the primary principal describes the movements of the primary principal based on a scenario.

16. The system of claim 11,

wherein while assessing the future states of the non-primary principal, the computation mechanism is configured to apply a dynamics model which describes the movements of the non-primary principal based on a scenario.

17. The system of claim 11,

wherein the mental-state model includes an “

alert”

state and a “

not-alert”

state; and

wherein the mental-state model specifies a first probability of transition from the “

alert”

state to the “

not-alert”

state and a second probability of transition from the “

not-alert”

state to the “

alert”

state.

18. The system of claim 11,

wherein the mental-state model includes a “

rational-decision”

state and an “

irrational-decision”

state; and

wherein the mental-state model specifies a first probability of transition from the “

rational-decision”

state to the “

irrational-decision”

state and a second probability of transition from the “

irrational-decision”

state to the “

rational-decision”

state.

19. The system of claim 11,

wherein a state of the primary or non-primary principal includes one or more of:

a position;

a velocity; and

a mental state of the person associated with the primary or non-primary principal.

20. The system of claim 11,

where the results include one or more of:

a probability of collision,

a predicted time of collision,

a predicted location of collision,

a predicted benefit of collision warning, and

an estimated prediction accuracy.

21. A computer system for facilitating assessment of collision between a primary principal and a non-primary principal for early warning, the system comprising:

a processor;

a memory; and

a specialized assessment mechanism comprising;

a data obtaining mechanism configured to obtain a current observation of the primary principal and non-primary principal;

a computation mechanism configured to assess one or more future states for the primary and non-primary principals, respectively, based on;

the current observation of the primary and non-primary principals,

a dynamics model of the primary principal, and

a mental-state model of a person associated with the primary principal.

22. The computer system of claim 21,

wherein while assessing the future states for the primary and non-primary principals, the computation mechanism is configured to perform sequential Bayesian filtering based on the current observation and past observations of the primary and non-primary principals, respectively.