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

How to eliminate AMCL particles using compass heading data?

asked 2018-07-10 19:58:35 -0600

achmad_fathoni gravatar image

After global_localization service called the initial particle is scattered everywhere with random heading. I have very good IMU compass say with 5 degrees error. My robot need to localize quickly so how to optimize particle filter by eliminating particles that differ more than 5 degrees from compass heading?

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2018-07-10 20:08:40 -0600

ahendrix gravatar image

You may be able to achieve this by setting the initial_pose_a and initial_cov_aa parameters described on the AMCL wiki page before starting the AMCL node. Obviously you'll need to choose a covariance (in radians) that corresponds to your compass's heading accuracy in order to generate the correct distribution of headings in the particle cloud.

edit flag offensive delete link more


Thank you for your answer. After few step from global_localization the particles start sticking together and make some groups. Often this group has different orientation to from another and only view match the compass heading. Can you give me a hint about how to filter the wrong particles groups?

achmad_fathoni gravatar image achmad_fathoni  ( 2018-07-11 01:35:50 -0600 )edit

I am ask this following question because initial_pose_a and initial_cov_a seems only cover initial particles.

achmad_fathoni gravatar image achmad_fathoni  ( 2018-07-11 01:37:09 -0600 )edit

I am not sure AMCL supports the general case of getting absolute heading from a sensor when doing online localization. I suggest you read the AMCL source code. You may need to modify the source code for AMCL if you want to filter particles by orientation during every update.

ahendrix gravatar image ahendrix  ( 2018-07-11 03:27:23 -0600 )edit

Question Tools



Asked: 2018-07-10 19:58:35 -0600

Seen: 339 times

Last updated: Jul 10 '18