Robot navigation - Localization

Process of figuring out where the robot is currently in a certain model of the environment, using sensor measurements and keeping track of the actual movement of the robot, to keep updating and computing the actual position.

Localization can be performed using different kinds of techniques, and heavily realies on using exteroception to match current preceptions of the world with past observations, updating the map after that. This matching usually is really hard, and usually wha are computed are a set of possible locations called poses.

Some techniques used for localization are:

• Iconic 20220228202048 : usage of an occupancy grid and direct measures of the environment.
• Feature based: use extracted features, similar to distinctive places 20220221202140

General localization approach.

• Create a local occupancy grid from past n readings.
• Every n readings, match the local grid with the global occupancy grid.
• Do this k times, one for every pose generated taking in consideration the translation and rotation (pose generation has to convert from a continuous range of k poses into a discrete finite one).
• Do the matching algorithm for selecting the best pose, taking in consideration uncertainty.
• When the best match is selected, set this as the current position and repeat the process.

Notes References

20210714190242 Robotics Basics - Navigation

20210514183815 INDEX - Robotics

References

(Murphy 2000) (Matariâc 2007)

Matariâc, Maja J. 2007. The Robotics Primer. Intelligent Robotics and Autonomous Agents Series. Cambridge, Mass: The MIT Press.

Murphy, Robin. 2000. Introduction to AI Robotics. Intelligent Robotics and Autonomous Agents. Cambridge, Mass: MIT Press.