outdoor localization with AMCL and Robot_Localization

asked 2022-07-16 01:18:35 -0500

gottfried gravatar image

updated 2022-07-19 03:56:04 -0500

Hi all,

I'm running a robot project that patrols outdoors. The road of robot patrol is higher than the surrounding roads and obstacles, which means that in some places, 3D lidar will be difficult to scan the obstacles.

The road is seems like a long elevated road without fence.

I use the following sensors: - 3d lidar - imu - gps and the following package - LIO_SAM for mapping - Pointcloud map to grid map package - Pointcloud_to_laserscan for convert pointcloud msgs to /scan - AMCL - Robot_localization - Move_base

After maping with LIO_SAM, I converted the 3D map into the occupation grid map, localization and navigation will be done in grid map.

In the areas with obvious obstacles, AMCL+Robot_Localizacation can complete the locating. But in the red circle area (see the pic ), the location is lost.

The following figure shows the node diagram and TF tree of the system(see the pic ).

What can I do to make the robot still locate accurately in the red circle position (that is, when the obstacle features are not obvious)?

The GPS I use claims that it can obtain centimeter level positioning, but I Not verified yet. (have tried utm in Robot_Localization )(see the pic ).

edit retag flag offensive close merge delete

Comments

Please include your configuration files, as well as sample sensor messages for each sensor input to the EKF. Your configuration is not clear from the description above. Are you running r_l for the odom->base_link transform and amcl for global localization (map->odom)? What does amcl do when you reach the location in question? Do the particles diverge?

Tom Moore gravatar image Tom Moore  ( 2022-08-26 03:32:22 -0500 )edit

Any update here?

Tom Moore gravatar image Tom Moore  ( 2023-03-09 08:08:36 -0500 )edit
add a comment