Linear Kalman Filter Deep Dive (and Target Tracking)

Overview

Class Central Tips

As a follow-on course to "Kalman Filter Boot Camp", this course derives the steps of the linear Kalman filter to give understanding regarding how to adjust the method to applications that violate the standard assumptions. Applies this understanding to enhancing the robustness of the filter and to extend to applications including prediction and smoothing. Shows how to implement a target-tracking application in Octave code using an interacting multiple-model Kalman filter.

Syllabus

Deriving the linear Kalman filter

Knowing how to derive the steps of the Kalman filter is important for understanding the assumptions that are made and to be able to re-derive the steps for different assumptions. This week, you will learn how to derive the steps and will gain insight into how the Kalman filter works.

Making the linear Kalman filter bulletproof

Last week, you learned the assumptions made when deriving the Kalman filter. What if these assumptions are not met correctly? What if numeric roundoff error causes failure? This week, you will learn how to solve problems with the standard Kalman filter.

Extensions and refinements to linear Kalman filters

The standard linear Kalman filter works well for state estimation, but can be extended to implement prediction and smoothing as well. Further, we can speed up the steps or even eliminate steps in some circumstances. This week, you will learn some extensions and refinements to linear Kalman filters.

Target-tracking application using a linear Kalman filter

A popular application of Kalman filters is to track (usually non-cooperating) targets. This week, you will learn how to implement standard and specialized Kalman filters suited for target tracking.