The Kalman Filter
Lifetime access to every lesson, the equations and the downloadable Python simulation. Your progress is saved to your account.
Every sensor lies a little. A GPS jumps around by metres, an encoder slips, a range sensor flickers. The Kalman filter is the standard way robots combine a model of how things move with noisy measurements to get an estimate that is better than either on its own.
In this course you will go from the intuition (two noisy guesses, one better answer) to the full matrix form, then build and tune a tracker for a moving rover in Python. The simulation lets you change the noise, the model and the filter settings and watch the estimate respond.
What you will be able to do
- Explain prediction, measurement and the Kalman gain in plain words
- Write the five Kalman filter equations from memory and know what every symbol means
- Tune Q and R and read the covariance to know how much to trust the estimate
- Build a constant-velocity tracker from scratch in Python with numpy
Before you start: School algebra, basic matrices (multiply, transpose, inverse) and a little Python.
Lessons
- The big idea: blending a guess with a measurement▶ VideoFree preview
- How it was invented and its first uses▶ VideoLocked
- Where it is used today▶ VideoLocked
- The equationsLocked
- Worked demo and tuningLocked
- Build your own, step by stepLocked
- Python simulation labLocked
- Practice problems and quizLocked