What do filter coefficient relationships mean?

Gray, John Edward; Smith-Carroll, A.S.; Murray, William J.

doi:10.1109/ssst.2004.1295615

Cited by 5 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While all filters have the same noise reduction ratios for position and velocity, they have different transient responses or bias depending on which threat model one uses. In matrix form, the predicted update is (note this model lumps maneuverability uncertainty in to the velocity component) X(k + 1) = FX(k) + T1Iw(k), (8) while the measurement model in matrix form is z(k + 1) = HX(k + 1) + n(k + 1); (9) where I K10 (10) [) = 1j (11) …”

Section: Constant Gain Filtersmentioning

confidence: 99%

“…In addition to bounding performance, the Jacobian of the noise reduction and transient performance ratios allow one to determine one of four possible relationship (Benedict-Bordner, Kalata [11], Continuous White Noise, and an unnamed one) between the filter coefficients, e.g. a = a (/3) as discussed in [9]. The other aspect of filter design is selection of the filter gains.…”

Section: Constant Gain Filtersmentioning

confidence: 99%

“…A mean squared cost function can be formed by a combination of noise reduction and filter response to an un-modeled term [9]. A normalized cost function (JN) (normalized means the mean squared error equals the normalized cost times the normalization constant N which for the steady state velocity cost function is (T2)).…”

Section: Constant Gain Filtersmentioning

confidence: 99%

See 2 more Smart Citations

Throwing Down The Gauntlet: A Discussion Of Techniques For Bounding Advanced Tracking Algorithm Performance

Gray

Smith-Carroll

2006

2006 9th International Conference on Information Fusion

View full text Add to dashboard Cite

For many applications of radar and sensor based filtering, simulations can not represent the sole estimate of performance, provide points where threats become engageable, or determine when to use weapons' platform based sensors effectively in an engagement, etc... No significant advances have been proposed to analytically characterize performance or at least bound performance of the Kalman filter other than the use of simple two or three state constant gain filters. This paper suggests methods for characterizing filter algorithms that can be used to bound the advanced tracking algorithms that are used in a single sensor or muli-sensor environment.

show abstract

Section: Constant Gain Filtersmentioning

confidence: 99%

Section: Constant Gain Filtersmentioning

confidence: 99%