Navigation system applications of sigma-point Kalman filters for nonlinear estimation and sensor fusion

• US 7,289,906 B2
• Filed: 04/04/2005
• Issued: 10/30/2007
• Est. Priority Date: 04/05/2004
• Status: Active Grant

• Current Assignee: Oregon Health & Science University
• Original Assignee: Oregon Health & Science University
• Inventors: van der Merwe, Rudolph, Wan, Eric A., Julier, Simon J.
• Primary Examiner(s): Zanelli; Michael J.
• Application Number:
  US11/099,433
• Publication Number:
  US 20050251328A1
• Time in Patent Office: 939 Days
• Field of Search: None
• US Class Current: 701/214
• CPC Class Codes: G01C 21/165 combined with non-inertial navigation instruments
• CPC Class Codes: G01S 19/49 whereby the further system is an inertial position system, e.g. loosely-coupled

Abstract

A method of estimating the navigational state of a system entails acquiring observation data produced by noisy measurement sensors and providing a probabilistic inference system to combine the observation data with prediction values of the system state space model to estimate the navigational state of the system. The probabilistic inference system is implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations. This technique achieves for the navigational state of the system an estimation accuracy that is greater than that achievable with an extended Kalman filter-based probabilistic inference system.

Claims

1. A method performed in an integrated navigation system to estimate the navigational state of an object, the navigational state characterized by random variables of a kinematics-based system state space model, the system state space model specifying a time evolution of the system and its relationship to sensor observations, comprising:

acquiring observation data produced by measurement sensors that provide noisy information about the navigational state, the measurement sensors including an inertial measurement unit (IMU) and a global positioning system (GPS) operating to provide information for estimating a set of navigational state components that include position, velocity, attitude, and angular velocity; and

providing a probabilistic inference system to combine the observation data with prediction values of the system state space model to estimate the navigational state, the probabilistic inference system implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations.

2. The method of claim 1, in which the navigational state component set includes bias and scale factor information associated with the IMU.

3. The method of claim 1, in which the navigational state component set includes phase and timing error information associated with the GPS.

4. The method of claim 1, in which the measurement sensors producing observation data constitute sensors of multiple orbital GPS satellites from which raw satellite signal transmissions propagate.

5. The method of claim 1, in which the random variable propagation technique of the probabilistic inference system used to estimate the navigational state includes:

using mean values and square root decomposition of a covariance matrix of a prior random variable to deterministically calculate a set of sigma-points and an associated set of weight values; and

propagating the calculated set of weighted sigma-points through the system state space model to produce a posterior set of weighted sigma-points used to calculate posterior statistics.

6. The method of claim 5, in which the calculation of posterior statistics entails the use of closed form functions of the propagated sigma-points and weight values.

7. The method of claim 5, further comprising using the calculated posterior statistics to give an approximately optimal update of the navigational state and its covariance by combining the prediction values of the system state space model and the sensor observations.

8. The method of claim 5, in which the probabilistic inference system used to estimate the navigational state is implemented with one of unscented Kalman filter (UKF) algorithms, central difference Kalman filter (CDKF) algorithms, square root version of UKF, square root version of CDKF, sigma-point particle filter (SPPF) algorithms, or Gaussian mixture sigma-point particle filter (GMSPPF) algorithms.

9. The method of claim 1, in which the observation data provide noisy information about the navigational state at a time that lags a current time because of sensor latency, communication delay, or other processing delay, and further comprising:

maintaining a cross correlation between a navigational state at a current time and the navigational state to which the latency lagged observation data correspond; and

using the cross correlation to optimally combine latency lagged observation data with the prediction values of the system state space model.

10. The method of claim 1, in which the GPS operates to receive raw satellite signal transmissions, and in which the received raw satellite signal transmissions are integrated directly with the IMU, thereby to form a tightly coupled integrated navigation system.

11. The method of claim 1, in which the position and velocity information provided by the GPS is integrated with the IMU, thereby to form a loosely coupled integrated navigation system.

12. An integrated navigation system for estimating the navigational state of an object, the navigational state characterized by random variables of a kinematics-based system state space model, the s stem state space model specifying a time evolution of the system and its relationship to sensor observations, comprising:

measurement sensors producing observation data in the form of noisy information about the navigational state, the measurement sensors including an integrated measurement unit (IMU) and a global positioning system (GPS) operating to provide information for estimating a set of navigational state components that include position, velocity, attitude, and angular velocity; and

a probabilistic inference system configured to combine the observation data with prediction values of the system state space model to estimate the navigational state, the probabilistic inference system implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations.

13. The system of claim 12, in which the observation data provide noisy information about the navigational state at a time that lags a current time because of sensor latency, communication delay, or other processing delay, and further comprising:

a computational subsystem operable to maintain a cross correlation between a navigational state at a current time and the navigational state to which the latency lagged observation data correspond; and

a computational subsystem operable to use the cross correlation to optimally combine latency lagged observation data with the prediction values of the system state space model.

14. The system of claim 12, in which the navigational state component set includes bias and scale factor information associated with the IMU.

15. The system of claim 12, in which the navigational state component set includes phase and timing error information associated with the GPS.

16. The system of claim 12, in which the measurement sensors producing observation data constitute sensors of multiple orbital GPS satellites from which raw satellite signal transmissions propagate.

17. The system of claim 12, in which the random variable propagation technique of the probabilistic inference system used to estimate the navigational state includes:

use of mean values and square root decomposition of a covariance matrix of a prior random variable to deterministically calculate a set of sigma-points and an associated set of weight values; and

propagation of the calculated set of weighted sigma-points through the system state space model to produce a posterior set of weighted sigma-points used to calculate posterior statistics.

18. The system of claim 17, in which the calculation of posterior statistics entails the use of closed form functions of the propagated sigma-points and weight values.

19. The system of claim 17, in which the random variable propagation technique further comprises use of the calculated posterior statistics to give an approximately optimal update of the navigational state and its covariance by combining the prediction values of the system state space model and the sensor observations.

20. The system of claim 17, in which the probabilistic inference system used to estimate the navigational state is implemented with one of unscented Kalman filter (UKF) algorithms, central difference Kalman filter (CDKF) algorithms, square root version of UKF, square root version of CDKF, sigma-point particle filter (SPPF) algorithms, or Gaussian mixture sigma-point particle filter (GMSPPF) algorithms.

21. The system of claim 12, in which the GPS operates to receive raw satellite signal transmissions, and in which the received raw satellite signal transmissions are integrated directly with the IMU, thereby to form a tightly coupled integrated navigation system.

22. The system of claim 12, in which the position and velocity information provided by the GPS is integrated with the IMU, thereby to form a loosely coupled integrated navigation system.

23. A method performed in a global positioning system (GPS) to estimate the navigational state of an object, the navigational state characterized by random variables of a kinematics-based system state space model, the system state space model specifying a time evolution of the system and its relationship to sensor observations, comprising:

acquiring observation data produced by measurement sensors that provide noisy information about the navigational state, the measurement sensors including a GPS operating to receive raw satellite signal transmissions and provide information for estimating a set of navigational state components including position and velocity; and

providing a probabilistic inference system to combine the observation data with prediction values of the system state space model to estimate the navigational state, the probabilistic inference system implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations.

24. The method of claim 23, in which the navigational state component set includes phase and timing error information associated with the GPS.

25. The method of claim 23, in which the measurement sensors producing observation data constitute sensors of multiple orbital GPS satellites from which raw satellite signal transmissions propagate.

26. The method of claim 23, in which the random variable propagation technique of the probabilistic inference system used to estimate the navigational state includes:

using mean values and square root decomposition of a covariance matrix of a prior random variable to deterministically calculate a set of sigma-points and an associated set of weight values; and

propagating the calculated set of weighted sigma-points through the system state space model to produce a posterior set of weighted sigma-points used to calculate posterior statistics.

27. The method of claim 26, in which the calculation of posterior statistics entails the use of closed form functions of the propagated sigma-points and weight values.

28. The method of claim 26, further comprising using the calculated posterior statistics to give an approximately optimal update of the navigational state and its covariance by combining the prediction values of the system state space model and the sensor observations.

29. The method of claim 26, in which the probabilistic inference system used to estimate the navigational state is implemented with one of unscented Kalman filter (UKF) algorithms, central difference Kalman filter (CDKF) algorithms, square root version of UKF, square root version of CDKF, sigma-point particle filter (SPPF) algorithms, or Gaussian mixture sigma-point particle filter (GMSPPF) algorithms.

30. A method of estimating a state of a system, the state characterized by random variables of a system state space model, the system state space model specifying a time evolution of the system and its relationship to sensor observations, comprising:

acquiring observation data produced by measurement sensors that provide noisy information about the state of the system at a time that lags a current time because of sensor latency, communication delay, or other processing delay;

providing a probabilistic inference system to combine the observation data with prediction values of the system state space model to estimate the state of the system, the probabilistic inference system implemented to include a realization of a Gaussian approximate random variable propagation technique performing deterministic sampling without analytic derivative calculations;

maintaining a cross correlation between a state of the system at a current time and the state to which the latency lagged observation data correspond; and

using the cross correlation to optimally combine latency lagged observation data with the prediction values of the system state space model.

31. The method of claim 30, in which the state of a system is a navigational state.