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

Any pkg which computes position from IMU

asked 2013-07-09 00:29:10 -0500

sai gravatar image

updated 2013-07-09 16:44:31 -0500


I was wondering if anyone can point me to ros package which can compute position from IMU.

I guess by integrating the linear acceleration, one could get that, but wanted to know if there is a package for that.

If not, could you tell how to efficiently write a program for that


edit retag flag offensive close merge delete



Are you sure, you want to do that? Unless you spent a lot of money on the IMU the results are probably very noisy.

dornhege gravatar image dornhege  ( 2013-07-09 01:43:44 -0500 )edit

I would like to give it as one of the input to Extended Kalman Filter(robot_pose_ekf pkg) along with other sources of odometry. Unfortunately, robot_pose_ekf pkg only takes 3D orientation from IMU, So I was wondering if its possible to use 3D position too..

sai gravatar image sai  ( 2013-07-09 04:17:27 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted

answered 2013-07-09 17:23:18 -0500

lindzey gravatar image

updated 2013-07-09 17:24:01 -0500

Strongly seconding @dornhege's comment.

The kalman filter is already doing the integration on the accelerations provided by the IMU, and using that to update its state estimate, so you're not going to get any additional accuracy by trying to spoof your own position estimates from the same data and feeding them to the ekf.

In order to improve your state estimation, I'd recommend doing a little reading on how kalman filters work (my favorite introductory explanation was in Probabilistic Robotics, I can't recommend a freely-available one off the top of my head), and then making sure that you have the right covariances for all of its inputs. If you want to improve it beyond that, you'll need to invest in better/different sensors.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2013-07-09 00:29:10 -0500

Seen: 1,480 times

Last updated: Jul 09 '13