Is the Kinect's IMU being used on the TurtleBot?

Kinect
imu

asked 2011-04-23 00:09:19 -0500

Pi Robot
4046 ●134 ●158 ●192 http://www.pirobot.org/

updated 2016-10-24 09:00:44 -0500

ngrennan
1 ●1 ●1

I seem to recall that the TurtleBot is using an IMU other than the accelerometer built into the Kinect. Is this true? And if so, what are the limitations of the Kinect's IMU that prevent it from being used instead?

Thanks!

edit retag flag offensive close merge delete

Comments

The Kinect only has a 2-axis accelerometer for tilt correct?

Eric Perko ( 2011-04-23 03:50:12 -0500 )edit

Ah, that makes sense if it's true. This page says it's a 3-axis: http://www.tomsguide.com/us/Microsoft-Kinect-Teardown-iFixit-Project-Natal-xbox-360,news-8640.html. And another page said the actual sensor is this: http://www.kionix.com/accelerometers/accelerometer-KXSD9.html which appears to be 3-axis. But when I run the kinect_aux package and echo the data on the /imu topic while manually moving the Kinect, the only non-zero numbers are the linear_accelerations, which vary on all three axes x, y, z. I think it is the orientations that are needed for the robot_pose_ekf node used with the Turtlebot. Is that correct? In other words, I think we need gyros, not an accelerometer, right?

Pi Robot ( 2011-04-23 15:27:04 -0500 )edit

add a comment

2

answered 2011-04-23 15:37:09 -0500

Eric Perko

8406 ●73 ●90 ●144 http://ericperko.com/

The Kinect only has an accelerometer, not a full IMU integrated into it. With the accelerometer, you could get an estimate of the pitch and roll of the Kinect relative to the ground (these would be based on the gravity vector) in addition to the instantaneous accelerations.

However, you would need a magnetometer to get an estimate of current yaw angle relative to magnetic north (or the strongest magnet near you... watch for magnetic door locks). Integrating a gyro could also give you the yaw orientation component, but integration can suffer from any bias that the gyro has on its measurements and, when even a small bias is compounded through repeated integration, could lead to inaccurate estimates for the yaw component of orientation.

If I remember correctly, the robot_pose_ekf package only cares about the orientation in the IMU message, not the accelerations or angular velocities. Can't say exactly what the Turtlebot setup uses though...

edit flag offensive delete link

Comments

Thanks Eric. I think that answers all my questions. I have a 9DOF Razor IMU that I am using on my fake Turtlebot (a mock turtle? :)) so I guess I'll stick with that.

Pi Robot ( 2011-04-23 15:43:34 -0500 )edit

add a comment

0

answered 2011-06-20 15:31:13 -0500