Localization

[lukicdarkoo]

We need to estimate the pose (position and orientation) of the robot. It is important to point that the localization can be local and global. The local one is subject to error accumulation, but the update rate is typically high and it is quite accurate (in the local frame). The global one is not subject to the error accumulation, but the update rate is typically low and it is quite inaccurate.

---

High-level tasks:

• Make sure the wheel odometry is very accurate
• Try different methods with a global localization:
  • UWB (Analyze UWB-based localization #1).
  • LiDAR and highly reflective beacons. It would be a simplified version of AMCL as we can only have 3 points.
  • Camera with ArUco markers. With known camera intrinsic and extrinsic parameters we can determine a pose of the ArUco markers placed on the robot.
• Implement EKF with the wheel odometry and the selected global localization.

---

Resources:

• Paper about EKF implementation: http://docs.ros.org/en/lunar/api/robot_localization/html/_downloads/robot_localization_ias13_revised.pdf
• Presentation by the maintainer of robot_localization package: https://www.youtube.com/watch?v=nfvvpYBAMww