ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

EKF for localization given a MAP

asked 2016-03-27 19:01:27 -0600

Karel Wosiack gravatar image

I'm starting to use the map obtained from de Gmapping with the objective of fusing this data with laser scan and odometry to make my robot localization using an EKF approach.

I search for an existing package without success, can you give me more information about this? Some examples or ideas?

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2016-03-28 12:11:46 -0600

al-dev gravatar image

updated 2016-03-28 16:28:51 -0600

Assuming that you specifically want to use an EKF-based approach and not just a particle filter with Monte Carlo localization like amcl, you could for instance use the EKF (or UKF) implementations in the robot_localization package to fuse your existing source of odometry with another odometry estimate made by a scan matcher like laser_scan_matcher.

Note that unless you have specific instructions not to use amcl, I would give it a try first since it is the most widely used localization tool for LIDAR+odometry robots with ROS.

Edit : If you want to use an existing map, you indeed need a scan-to-map matcher. Then robot_localization can help fuse the estimate of the scan-to-map matcher with your odometry using an EKF. But I am afraid I don't know about individual ROS packages providing a scan-to-map matching feature in isolation, you would have to look at how it is implemented in the various localization/SLAM packages out there.

edit flag offensive delete link more


Yes, I specifically want to use an EKF-based approach. I want to do localization as made by the AMCL but using one EKF approach. I want to use the map information too, and as I can see the laser_scan_matcher doesn't use it. Do you know any existing approach that does it?

Karel Wosiack gravatar image Karel Wosiack  ( 2016-03-28 16:49:27 -0600 )edit

unless you are not allowed to use AMCL, I would recommend using AMCL with an odometry source and then, if you want, fusing them together using an EKF like robot_localization (I know this is a year old, but I wanted to answer this for anyone who looks at this question later)

tb12 gravatar image tb12  ( 2017-02-27 13:47:32 -0600 )edit

Question Tools



Asked: 2016-03-27 18:52:44 -0600

Seen: 1,627 times

Last updated: Mar 28 '16