State estimation using motion context and multiple input observation types

• US 8,615,253 B2
• Filed: 06/03/2011
• Issued: 12/24/2013
• Est. Priority Date: 06/03/2011
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: MacGougan, Glenn Donald, Marti, Lukas M., Mayor, Robert
• Primary Examiner(s): Bhattacharya, Sam
• Application Number:
  US13/153,329
• Publication Number:
  US 20120309411A1
• Time in Patent Office: 935 Days
• Field of Search: 455/456.1, 455/456.3, 340/357.23, 340/357.25, 340/357.63
• US Class Current: 455/456.1
• CPC Class Codes: G01S 5/0258 by combining or switching between measurements derived from different systems
• CPC Class Codes: G01S 5/0294 Trajectory determination or predictive filtering, e.g. target tracking or Kalman filtering
• CPC Class Codes: H04W 64/006 with additional information processing, e.g. for direction or speed determination

Abstract

Techniques for estimating the current state (e.g., position, velocity) of a mobile device based on motion context and multiple input observation types are disclosed. In some implementations, an Extended Kalman Filter (EKF) formulation is used to combine multiple input observations received from a variety of sources (e.g., WiFi, cell, GPS) to compute a minimum error state estimate. In some implementations, the EKF is updated using position estimates from an active cell and/or a candidate active cell during a cell-hopping event.

Claims

1. A method performed by a mobile device, comprising:

selecting a state observation of the mobile device from a set of observation types including network-derived observations and satellite-derived observations, where the selected observation is input to a state estimator that is configured for receiving network-derived and satellite-derived observations;

determining that a motion context for the mobile device has changed from moving to static or from static to moving;

determining one or more dynamic model parameters based on the changed motion context; and

estimating a state of the mobile device using the state estimator based on the selected observation type and determined dynamic model parameters,

where the method is performed by one or more hardware processors.

2. The method of claim 1, where network-derived observations are estimated positions derived from at least one of WiFi and cellular networks.

3. The method of claim 1, where the satellite-derived observations are estimated positions or velocities derived from a global navigation satellite system (GNSS).

4. The method of claim 1, where determining a motion context of the mobile device includes determining that the mobile device is moving and the type of movement.

5. The method of claim 1, where estimating a state of the mobile device includes estimating a position or velocity of the mobile device using an estimation filter that includes a dynamic model that is adjusted by the dynamic model parameters to control the uncertainty of growth of the position and velocity estimates.

6. The method of claim 5, where the estimation filter is an extended Kalman filter (EKF).

7. The method of claim 6, where the dynamic model is adjusted by adding additional process noise in a variance-covariance prediction step of the EKF, where the magnitude of the additional process noise depends on the motion context.

8. The method of claim 7, where the additional process noise is generated by a Gauss-Markov process.

9. The method of claim 5, where determining a motion context of the mobile device further comprises:

determining that the mobile device is static; and

updating the EKF using a zero-velocity update.

10. A method performed by a mobile device operating in a wireless network, the method comprising:

determining a motion context of the mobile device;

tuning a state estimator running on the mobile device based on the determined motion context;

detecting a cell-hopping event; and

updating the state estimator operating on the mobile device based on the determined motion context and the detected cell-hopping event,

where the method is performed by one or more hardware processors.

11. The method of claim 10, where the estimator is an extended Kalman filter (EKF).

12. The method of claim 11, further comprising:

updating the EKF with a first position estimate from an active cell; and

updating the EKF with a second position estimate from the candidate active cell.

13. The method of claim 11, further comprising:

determining a dwell time of the mobile device in the active cell;

if the dwell time exceeds a threshold,

updating the EKF with a first position estimate from the active cell at the same update epoch as the position update from the candidate active cell;

updating the EKF with a second position estimate from the candidate active cell; and

if the dwell time does not exceed the threshold,

updating the EKF with a position estimate from the candidate active cell.

14. A system comprising:

one or more processors;

memory storing instructions, which, when executed by the one or more processors, causes the one or more processors to perform operations comprising;

selecting a state observation of a mobile device from a set of observation types including network-derived observations and satellite-derived observations, where the selected observation is input to a state estimator that is configured for receiving network-derived and satellite-derived observations;

determining that a motion context for the mobile device has changed from moving to static or from static to moving;

determining one or more dynamic model parameters based on the changed motion context; and

estimating a state of the mobile device using the state estimator based on the selected observation type and determined dynamic model parameters,

where the method is performed by one or more hardware processors.

15. The system of claim 14, where network-derived observations are estimated positions derived from at least one of WiFi and cellular networks.

16. The system of claim 14, where the satellite-derived observations are estimated positions or velocities derived from a global navigation satellite system (GNSS).

17. The system of claim 14, where determining a motion context of the mobile device includes determining that the mobile device is moving and the type of movement.

18. The system of claim 14, where estimating a state of the mobile device includes estimating a position or velocity of the mobile device using an estimation filter that includes a dynamic model that is adjusted by the dynamic model parameters to control the uncertainty of growth of the position or velocity estimates.

19. The system of claim 18, where the state estimator is an extended Kalman filter (EKF).

20. The system of claim 19, where the dynamic model is adjusted by adding additional process noise in a variance-covariance prediction step of the EKF, where the magnitude of the additional process noise depends on the motion context.

21. The system of claim 20, where the additional process noise is generated by a Gauss-Markov process.

22. The system of claim 20, where determining a motion context of the mobile device further comprises:

determining that the mobile device is static; and

updating the EKF using a zero-velocity update.

23. A system comprising:

one or more processors;

memory storing instructions, which, when executed by the one or more processors, causes the one or more processors to perform operations comprising;

determining a motion context of a mobile device;

tuning a state estimator running on the mobile device based on the determined motion context;

detecting a cell-hopping event; and

updating the state estimator operating on the mobile device based on the determined motion context and the detected cell-hopping event.

24. The system of claim 23, where the state estimator is an extended Kalman filter (EKF).

25. The system of claim 24, where the operations further comprise:

updating the EKF with a first position estimate from the active cell; and

updating the EKF with a second position estimate from the candidate active cell.

26. The system of claim 23, where the operations further comprise:

determining a dwell time of the mobile device in the active cell;

if the dwell time exceeds a threshold,

updating the EKF with a first position estimate from the active cell at the same update epoch as the position update from the candidate active cell;

updating the EKF with a second position estimate from the candidate active cell; and

if the dwell time does not exceed the threshold,

updating the EKF with a position estimate from the candidate active cell.

27. A method performed by a mobile device, comprising:

receiving a state update;

determining if an elapsed time since a last state update is less than a predetermined time period; and

if the elapsed time is less than a predetermined time period;

providing a state estimate with a first state estimator, and

re-seeding a second state estimator with parameters from the first state estimator,

where the second estimator is configured based on a long-term static assumption;

if the elapsed time is greater than a predetermined time period;

updating the second state estimator with the state update;

performing a static hypothesis test and an innovation test, and

if the hypothesis and innovation tests pass,

copying motion context data from the first estimator to the second estimator; and

re-seeding the first state estimator with parameters from the second state estimator,

where the method is performed by one or more hardware processors.