The Unscented Kalman Filter (UKF)
Lifetime access to every lesson, the equations and the downloadable Python simulation. Your progress is saved to your account.
Pay securely with UPI, cards, netbanking or wallets through Razorpay. The course unlocks instantly.
By enrolling you agree to our Terms, Refund policy and Privacy policy.
The EKF needs Jacobians and can be badly wrong when models curve sharply. The unscented Kalman filter replaces derivatives with a small set of carefully chosen sample points, the sigma points, which are pushed through the real nonlinear model. It is usually more accurate than the EKF at similar cost and much easier to apply to complicated models.
You will learn the idea, the history, the exact equations and how to pick the tuning parameters, then build a UKF that tracks a pendulum and race it against an EKF.
What you will be able to do
- Explain the unscented transform and why it beats linearisation
- Generate sigma points and weights with the scaled (α, β, κ) scheme
- Implement a UKF from scratch with Cholesky decomposition
- Compare EKF and UKF accuracy on a strongly nonlinear system
Before you start: Courses 1 and 2, matrix square roots (Cholesky) will be explained.
Lessons
- The big idea: push points, not slopes▶ VideoFree preview
- How it was invented and its first uses▶ VideoLocked
- Where it is used today▶ VideoLocked
- The equations▶ VideoLocked
- Worked demo and tuning▶ VideoLocked
- Build your own, step by step▶ VideoLocked
- Python simulation labLocked
- Practice problems and quizLocked