Skip to main content
#slam#localization#sensor-fusion#kalman-filter#particle-filter

SLAM, Part 2

A second set of study notes reviewing the stages of simultaneous localization and mapping.

I wrote this post to review the stages of a SLAM pipeline.

#What is SLAM?

SLAM stands for simultaneous localization and mapping. A robot estimates its own pose while building a map.

  1. If a moving robot knows its pose, it can build a map.
  2. If it has a map, it can estimate its pose. SLAM solves these two problems together.

#The SLAM pipeline

The pipeline has three main stages.

  1. Receive sensor data.
  2. Estimate the robot’s pose.
  3. Build a map.

#Sensor data

For more on the sensors, see the first SLAM article. 🤗

#Estimating the robot’s pose

The robot estimates its pose from the observations it receives. A walkthrough of the principle: MATLAB’s SLAM explanation The video gives an accessible introduction.

Processing sensor data Raw measurements contain errors. Typical processing steps include (1) noise reduction, (2) data registration, and (3) interpolation. The details depend on the sensor. Vision, LiDAR, and IMU data each merit a separate discussion.

Robot pose estimation Several estimation algorithms can be used for this stage.

  1. EKF (Extended Kalman Filter)
  2. UKF (Unscented Kalman Filter)
  3. PF (Particle Filter)
  4. Other estimators.

I plan to cover these methods in a later post.


Connect