Interacting multiple bias model filter system for tracking maneuvering targets

• US 5,325,098 A
• Filed: 06/01/1993
• Issued: 06/28/1994
• Est. Priority Date: 06/01/1993
• Status: Expired due to Fees

• Current Assignee: the united states of america as represented by the secretary of the navy
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Rice, Theodore R., Watson, Gregory A., Blair, William D.
• Primary Examiner(s): Hellner, Mark
• Application Number:
  US08/069,818
• Time in Patent Office: 392 Days
• Field of Search: 342/95, 342/96, 342/101, 342/195
• US Class Current: 342/95
• CPC Class Codes: G01S 13/726 Multiple target tracking

Abstract

A filtering system used in the tracking of a maneuvering target is provid A first filter estimates a partial system state at a time k in terms of target position measurements. A plurality of second filters are each provided with an acceleration model hypothesis from a prior time (k-1) free of position and velocity constraints. Each second filter generates an acceleration estimate at time k and a likelihood at time k that the acceleration model hypothesis is correct. The likelihoods from the second filters are summed to generate a probability vector at time k. A third interaction mixing filter generates the acceleration model hypotheses from prior time (k-1) using the probability vector from prior time (k-1) and the acceleration estimates from prior time (k-1). The third filter also provides an error covariance to the first filter to reflect the uncertainty in the acceleration model hypotheses from prior time (k-1). A probabilistic acceleration estimate for time k is formed as a sum of each of the acceleration estimates associated with each of the acceleration model hypotheses multiplied by a corresponding probability from the probability vector. The partial system state estimate and the probabilistic acceleration estimate are summed to generate a complete system state estimate at time k in terms of position, velocity and acceleration of the maneuvering target.

Claims

1. A filtering system used in tracking of a maneuvering target comprising:

a target tracking sensor providing target position measurements;

first filter means for determining a partial system state estimate at a time K in terms of said measurements;

means for generating an error difference at time K between said partial system state estimate and said measurements;

bias filter means provided with prior acceleration model hypotheses from a time (K-1) free of position and velocity constraints and said error difference for generating acceleration estimates at time K and likelihoods at time K that said prior model hypotheses are correct;

updating means summing said likelihoods from said bias filter means for generating a probability vector at time K defining probability associated with each said prior acceleration model hypothesis;

interaction mixing filter means for generating each said prior acceleration model hypothesis using said probability vector from prior time (K-1) and said acceleration estimates from said prior time (K-1) as a probabilistic sum of said acceleration estimates from prior time (K-1) multiplied by said probability vector from prior time (K-1), means associated with said interaction mixing filter means for supply of an error covariance to said first filter means to reflect therein an uncertainty with respect to each said prior acceleration model hypothesis;

bias estimator means for generating a probabilistic acceleration estimate from each of said acceleration estimates and said probability vector at time K as a sum of each of said acceleration estimates associated with each of said prior acceleration model hypotheses multiplied by a corresponding probability from said probability vector; and

means summing said partial system state estimate and said probabilistic acceleration estimate for generating a complete system state estimate at time K defining position, velocity and acceleration of said maneuvering target.

2. A system as in claim 1 wherein said bias filter means includes constant acceleration filters.

3. A system as in claim 1 wherein said target tracking sensor further provides target velocity measurements.

4. A system as in claim 1 wherein said first filter means and each said bias filter means comprises a Kalman filter network.

5. A system as in claim 1 wherein said system state estimate at time k is a feedback input to said target tracking sensor.

6. A system as in claim 1 further comprising means for displaying said system state estimate at time k.

7. In a target tracking system having a target tracking sensor, bias-free filter means rendered operative by target measurement data form the sensor for supplying target position and velocity estimates and an interacting multiple model filter network operatively connected to the sensor for correcting said target position and velocity estimates, including:

an interaction mixer providing prior model hypotheses and means supplying error covariance from the interaction mixer to the bias-free filter means for modifying operation thereof reflecting uncertainty of the prior model hypotheses.

8. The system as defined in claim 7 including an input summer connected to the sensor through which the target measurement data is supplied to the bias-free filter means and the interacting multiple model filter network, and time delay means connected to the bias-free filter means through which prior target measurement data is supplied to the interacting multiple model filter network from which the prior model hypotheses is derived by the interaction mixer.

9. The system as defined in claim 8 wherein the interacting multiple model filter network further includes:

bias filters receiving the target measurement data from the sensor through the input summer to provide target acceleration estimates, a model probability updater receiving likelihood data from the bias filters to provide a probability vector fed to the interaction mixer and combiner means receiving the target acceleration estimates from the bias filters and the probability vector from the updater for providing an overall target state estimate.

10. The system as defined in claim 7 wherein the interacting multiple model filter network further includes:

bias filters receiving the target measurement data from the sensor to provide target acceleration estimates, a model probability updater receiving likelihood data from the bias filters to provide a probability vector fed to the interaction mixer and combiner means receiving the target acceleration estimates from the bias filters and the probability vector from the updater for providing an overall target state estimate.