ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
1

How to evaluate AMCL localization performance

asked 2012-08-28 03:14:46 -0500

Ibraim gravatar image

Hi all,

I'm using amcl for localization in a mobile robot. Now succesfully I have it working on the robot. My idea is to use it with a laser scan and I will be modifying the readings so the information passed to the localization algorithm is somehow filtered.

What I want to do now is to evaluate the performance of one filter vs. the original data, is there any information that amcl output so I can do it?. I think I can use the spread of the particles, but seeing from the execution this doesn't always reflect the accuracy of the localization, as sometimes the variance is small but the robot is not in its right orientation.

Other way to ask the question would be: how can I compare odometry vs. laser base localization using amcl algorithm?

Thanks for your thoughts and ideas,

Ibraim

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted
1

answered 2013-02-08 14:58:35 -0500

HammyG gravatar image

Have you considered using stage simulation? That way you have the ground truth pose (exactly where the robot is in the simulated world) which you can compare against the amcl pose.

For example, you can compare the euclidean distance between the two poses after moving the robot on a fixed path. Run this for both laser inputs, and whichever has the lowest (average!) distance is presumably the better one.

edit flag offensive delete link more

Question Tools

1 follower

Stats

Asked: 2012-08-28 03:14:46 -0500

Seen: 1,622 times

Last updated: Feb 08 '13