Courses / theory

Kalman Filters

intermediate~15 hours16 lessonsEnvironment: pyodide

Prerequisites: maths-for-robotics, kinematics-mobile-robots

Start: State, belief, and uncertaintyPractice questions

Modules

Module 1 — Thinking in probabilities
  1. 1State, belief, and uncertainty
  2. 2Bayes rule for robots
  3. 3Gaussians everywhere
  4. 4Widget: belief updater
Module 2 — The 1D Kalman filter
  1. 1Predict step
  2. 2Update step
  3. 3Tuning Q and R
  4. 4Widget: interactive 1D Kalman filter
Module 3 — Multivariate Kalman filters
  1. 1State vectors and covariance matrices
  2. 2The full KF equations
  3. 3Tracking position and velocity
  4. 4Project: KF for noisy robot odometry
Module 4 — Nonlinear filters
  1. 1Why linearity breaks (and EKF fixes it)
  2. 2The Extended Kalman Filter
  3. 3UKF and particle filters at a glance
  4. 4Where these live in ROS (robot_localization, AMCL)

Textbooks

  • Kalman and Bayesian Filters in Python — Roger Labbefree-book
    Free Jupyter book on GitHub
  • Probabilistic Robotics — Thrun, Burgard & Foxbook

Video resources

  • Kalman filter lectures — Cyrill Stachniss