Ask Your Question
0

robot_pose_ekf wrong output with "good input"

asked 2013-06-06 05:29:03 -0500

jep31 gravatar image

updated 2013-06-23 22:09:20 -0500

I'm not be able to understand why I get a wrong pose from robot_pose_ekf. I'm using a kinect on a table and connected to the workstation to find the turtlebot position orientation and publish the result in /vo. Kinect is tracking a pattern marker and find a pose. I want to use odometry and imu to have a better result and have a pose when kinect can't see the marker.

My issue is /vo, /odom, and imu/data has similar orientation but ekf find something very different and I can't understand why.

Here is my tf tree to see how I work:image description

/vo:

header:    seq: 479   stamp: 
    secs: 1370526479
    nsecs: 1198623   frame_id: odom child_frame_id: base_footprint pose:   pose: 
    position: 
      x: 0.109848357611
      y: 0.0184851177529
      z: 0.00108856573807
    orientation: 
      x: -0.0155068317083
      y: 0.00236792210117
      z: -0.726064768505
      w: 0.687447367477   covariance: [0.001, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.001, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.001, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 10000.1, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 10000.1, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-06]

/odom

header: 
  seq: 5144
  stamp: 
    secs: 1370526478
    nsecs: 985132932
  frame_id: odom
child_frame_id: base_footprint
pose: 
  pose: 
    position: 
      x: 0.122778090562
      y: 0.0180749352772
      z: 0.0
    orientation: 
      x: 0.0
      y: 0.0
      z: -0.754709580223
      w: 0.656059028991
  covariance: [0.001, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.001, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1000000.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1000000.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1000000.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1000.0]

/imu/data:

header: 
  seq: 5154
  stamp: 
    secs: 1370526479
    nsecs: 318391084
  frame_id: gyro_link
orientation: 
  x: 0.0
  y: 0.0
  z: -0.641532116852
  w: 0.767096175878
orientation_covariance: [1000000.0, 0.0, 0.0, 0.0, 1000000.0, 0.0, 0.0, 0.0, 1e-06]

and robot_pose_ekf/odom:

header: 
  seq: 3340
  stamp: 
    secs: 1370526479
    nsecs: 618319034
  frame_id: odom
pose: 
  pose: 
    position: 
      x: 0.0979321075308
      y: -0.0259189421082
      z: -0.00496891294622
    orientation: 
      x: 3.68249858201e-06
      y: -4.82461122176e-05
      z: 0.970055184809
      w: -0.242884614753
  covariance: [5.493486503712613e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 5.493486503712613e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0987278074026108e-06, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 10.877878440724317, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 10.877878440724317, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 5.493333772345698e-10]

And how I construct the odometry msg:

 void readAndSendVo()
  {
    tf::TransformListener lisVo;
    tf::StampedTransform transformVo;
    nav_msgs::Odometry vo;
    std ...
(more)
edit retag flag offensive close merge delete

Comments

just a note: 'ntpdate -q' does a query only, and does not set the time. you need to 'sudo ntpdate <master IP>' in order to sync

weiin gravatar image weiin  ( 2013-06-06 16:31:15 -0500 )edit

some troubleshooting: remove /imu and /vo as ekf inputs. ie see if only /odom topic going through ekf will affect the output. then add /vo. then add /imu. (i'm suspecting /imu is creating the error.. just a hunch)

weiin gravatar image weiin  ( 2013-06-06 16:34:07 -0500 )edit

I did your new sync with same result. I can see the msg vo, imu and odom sensor activated but often vo goes deactivated and return activated quickly. I have also the msg: "Robot pose ekf diagnostics discovered a potential problem"and "No devices connected.... waiting for devices to be connected"

jep31 gravatar image jep31  ( 2013-06-06 21:08:15 -0500 )edit

Often imu orientation is very different from odom orientation, is that normal ? And if no, how can I fix it ?

jep31 gravatar image jep31  ( 2013-06-06 21:18:18 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted
0

answered 2013-07-04 17:14:59 -0500

weiin gravatar image

Not sure what imu you are using. If you are using one that uses magnetometer (compass) to get orientation data, it might be affected by magnetic fields around it. Check this by placing your robot stationary and see if /imu/data orientation values drift.

edit flag offensive delete link more

Comments

Yes It does, position is moving in stationary position because velocity has some noise sometimes. I prefer not to use it for now. I'm writing a new kalman filter because this package is not enable to use global sensor anyway.

jep31 gravatar image jep31  ( 2013-07-07 21:01:10 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

1 follower

Stats

Asked: 2013-06-06 05:29:03 -0500

Seen: 769 times

Last updated: Jul 04 '13