Particle Filter in Robotics

What is Particle Filter in Robotics?

A particle filter (Monte Carlo localisation) represents a robot's belief about its pose as a cloud of thousands of weighted samples (particles). Each particle is a hypothesis about where the robot might be; their weights encode how likely each one is given recent measurements.

How It Works

At each step the filter does three things. Predict: every particle is moved according to the motion command plus random noise — making the cloud spread. Weigh: every particle's weight is updated by how well its predicted sensor readings match the actual readings. Resample: particles with low weight are discarded and high-weight ones are duplicated, focusing the cloud on the most likely locations. Over time, the cloud collapses to the true robot pose.

Real-World Example

ROS2's AMCL package (Adaptive Monte Carlo Localisation) is a particle filter used in thousands of warehouse robots. Roomba uses a particle filter to recover localisation after being picked up. Self-driving cars often run particle filters in parallel with Kalman-family filters as a backup.

Why It Matters for Robotics

Particle filters handle situations where Kalman and EKF fail: highly nonlinear maps, kidnapped-robot recovery, and ambiguous corridors. They are computationally heavier but rock-solid in the messy real world. Every senior robotics engineer should be able to explain and tune one.

Try It Yourself

Open Nav2 docs and run the AMCL tutorial in Gazebo. Pick the robot up mid-simulation (the "kidnapped robot" test) and watch the particle cloud explode and re-converge — a visceral demonstration of why particle filters matter.

Quick Quiz

🧠
Quick Quiz
3 questions
1.A particle filter represents belief as:
A.One Gaussian
B.Many weighted sample hypotheses
C.A single number
D.A neural network only
2.The "resample" step in a particle filter:
A.Adds new sensors
B.Duplicates high-weight particles and drops low-weight ones
C.Increases battery
D.Reduces dimension
3.Particle filters beat EKFs at:
A.Battery life
B.Multi-modal and highly nonlinear distributions
C.Single-Gaussian small problems
D.Memory usage

Further Reading

• Kalman Filter
• Extended Kalman Filter
• Amcl
• Slam
• Sensor Fusion

Ask R2 About This

Open the R2 Co-pilot (press ⌘K anywhere on R2BOT) and ask: "Explain Particle Filter in Robotics for a Class 9 student in India, with one real-world Indian example." You'll get a tailored, sourced answer in seconds.