AMCL (Adaptive Monte Carlo Localisation)
314 words · 2 min read · 1 source
AMCL is ROS2's standard particle-filter localiser. Given a known map, it tracks where the robot is using laser scans.
AMCL (Adaptive Monte Carlo Localisation) is a particle-filter algorithm that localises a robot in a previously built map using LIDAR (or sonar) data. It is the default localiser in ROS2 Nav2 and runs on thousands of mobile robots worldwide.
What is AMCL (Adaptive Monte Carlo Localisation)?
AMCL (Adaptive Monte Carlo Localisation) is a particle-filter algorithm that localises a robot in a previously built map using LIDAR (or sonar) data. It is the default localiser in ROS2 Nav2 and runs on thousands of mobile robots worldwide.
How It Works
AMCL maintains a population of particle hypotheses about the robot's pose. At each control step it predicts each particle forward using odometry plus noise, then weights it by how well its predicted scan matches the actual LIDAR scan. Resampling concentrates particles on high-likelihood poses. The 'adaptive' part dynamically adjusts the number of particles based on convergence — fewer when localised, more when uncertain (e.g., after the robot is kidnapped).
Real-World Example
Almost every Nav2-based robot uses AMCL out of the box. GreyOrange Butlers, Robotis TurtleBots, hospital delivery robots, and Indian academic mobile-robotics projects all rely on it. AMCL handles minor map changes (people walking by) gracefully.
Why It Matters for Robotics
AMCL is the first localiser most ROS2 users touch. Knowing how to tune its parameters — number of particles, motion model noise, laser model — is essential for any real deployment. AMCL bugs (poor convergence, slow recovery) are a daily occurrence in production.
Try It Yourself
Launch the Nav2 TurtleBot3 demo in Gazebo. Use RViz to publish an initial pose estimate and watch the particle cloud collapse. Then teleport the robot to a random spot and use the '2D Pose Estimate' tool to recover — feel AMCL converge.
Quick Quiz
Quick Quiz
3 questions
1.AMCL is based on which underlying algorithm?
2.AMCL requires:
3.The "adaptive" in AMCL refers to:
Further Reading
Ask R2 About This
Open the R2 Co-pilot (press ⌘K anywhere on R2BOT) and ask: "Explain AMCL (Adaptive Monte Carlo Localisation) for a Class 9 student in India, with one real-world Indian example." You'll get a tailored, sourced answer in seconds.
🐍 Python Playground · runs in your browser
Editor · 15 lines
Output
Press ▶ Run to execute. First run downloads Python (~6MB) — only happens once per page.
Powered by Pyodide · Python in WebAssembly · no server required.
Next: the course path
Five free courses, from zero to your own robot design
Short videos with a quiz after each one, a final activity and a certificate per course.
See the course path →Plain English. Free. Unsubscribe in one click.
Keep going
Particle Filter in Robotics
Particle filters represent uncertainty with thousands of weighted samples — perfect for nonlinear, multi-modal…
ConceptSLAM
SLAM is the technique a robot uses to build a map of an unfamiliar place — while figuring out where it is on t…
ConceptLIDAR SLAM in Robotics
LIDAR SLAM uses laser-scan data to map an environment and localise a robot. The gold standard for accuracy in …
Last updated · 2026-05-21
Spotted something off? Report an error →