v0.2.2

Try-catch statements have been put around state update functions (due to possible rounding error in lowrankdowndate operations). In such cases, covariance updates are ignored and allowed to increase. This has been tested on industrial applications.

v0.2.1

Wraps covariance update step in a try-catch statement to skip step if roundoff error results in a non positive-definite matrix.

v0.2.0

New outlier suppression scheme works directly on Kalman gain, which also affects the posterior covariance (reducing how much a measurement affects the posterior as it is being suppressed). This is not a substitute for logic outside the Kalman filter to deal with high-conflict scenarios (where the innovation value is far outside the expected distribution). It does however, greatly mitigate the risk of a single measurements from corrupting the state before such logic can be applied.

v0.1.2

Improved performance, simpler API, verified on a collection of industrial applications

v0.1.1

Proven, used in various industrial applications.

0.1.0

Direct port from personal codebase