Ask Your Question

How to integate the package with sensors such as lidar or cameras? [closed]

asked 2018-11-22 23:46:04 -0600

gaochao_hit gravatar image

I am little confused with the robot_localization package about how to integrate it with laser. If I have an IMU, an odometry and a laser,how should I input the pose estimated by registration of cloud points ?

edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by gaochao_hit
close date 2018-12-05 03:22:35.326635

1 Answer

Sort by ยป oldest newest most voted

answered 2018-11-26 04:21:13 -0600

Tom Moore gravatar image

The robot_localization package was not named well, and so I can understand the confusion of what it actually does. Its primary purpose is to perform multi-sensor fusion. See the wiki for more details:

The IMU and odometry data can be readily fused into one of the state estimation nodes in r_l. For the laser data, you're going to need a node that performs scan-to-map matching (e.g., amcl). You would then fuse the output of that node into, e.g., the r_l EKF. Careful, though: amcl uses your robot's odometry to project the current hypothesis set forward in time, so by fusing it with your odometry and IMU data, you'll be effectively including your wheel odometry data twice.

A better model for using amcl is probably to fuse the wheel encoder and IMU data alone in an EKF that generates the odom->base_link transform (set the world_frame parameter to odom), and then let amcl use that as its odometry information. amcl would then generate the map->base_link transform.

edit flag offensive delete link more


Thank you for your answer. In fact, I am now studying on global localization using 3D lidar. I have achived global localization using NDT and IMU in UKF framework by myself.But there arise one problem that the localization output often suffer abrupt change, which will do harm to navigation.

gaochao_hit gravatar image gaochao_hit  ( 2018-11-27 05:14:04 -0600 )edit

So I want to know whether the robot_localization package can improve the smoothness of the localization result and would you please offer some suggestions ?

gaochao_hit gravatar image gaochao_hit  ( 2018-11-27 05:15:48 -0600 )edit

Sounds like you need to use a two-tiered system, which is common. It's not a GPS, but the ideas are the same. See this:

Tom Moore gravatar image Tom Moore  ( 2018-11-27 07:34:31 -0600 )edit

Read the section called "Notes on Fusing GPS Data."

Tom Moore gravatar image Tom Moore  ( 2018-11-27 07:34:53 -0600 )edit

Thanks for your help.

gaochao_hit gravatar image gaochao_hit  ( 2018-11-27 21:15:25 -0600 )edit

Hi @Tom Moore, do you have any document about the first combination (Use EKF to fuse AMCL result with Odometry result) ? I think using odometry 2 times like that is not ok, but I don't know why. I want to figure out this, thank you!

N.N.Huy gravatar image N.N.Huy  ( 2021-05-16 05:20:12 -0600 )edit

Question Tools


Asked: 2018-11-22 23:42:13 -0600

Seen: 654 times

Last updated: Nov 26 '18