Courses / theory

SLAM: Mapping and Localization

advanced~16 hours13 lessonsEnvironment: pyodide

Prerequisites: python3, kinematics-mobile-robots, kalman-filters

Start: Why maps must be built, not givenPractice questions

Modules

Module 1 — The mapping problem
  1. 1Why maps must be built, not given
  2. 2Occupancy grids and log-odds
  3. 3Project: an occupancy grid mapper (known poses)
Module 2 — Scan matching
  1. 1The alignment problem: where did I move?
  2. 2ICP: iterative closest point
  3. 3Widget: the scan matcher
Module 3 — Graph SLAM
  1. 1Poses as a graph: nodes, edges, constraints
  2. 2Loop closure: recognizing where you've been
  3. 3Widget: watch the graph relax
  4. 4Project: 2D graph SLAM from scratch
Module 4 — SLAM in practice
  1. 1slam_toolbox: the production mapper, decoded
  2. 2Mapping session craft: driving for good maps
  3. 3Capstone: the full pipeline on recorded scans

Textbooks

  • Probabilistic Robotics — Thrun, Burgard & Foxbook