ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
2

Can a low cost IMU compensate for wheel slip induced heading errors?

asked 2014-06-27 10:14:17 -0600

ChrisL8 gravatar image

I have a two wheeled differential drive robot with a Kinect, similar in shape and form to a TurtleBot. The odometry data provides for beautiful straight lines, but when turning there is a tremendous amount of wheel slip on some surfaces, but little or none on others, making it impossible to predict. The result is very strangely shaped rooms.

I have already experimented with compasses (magnetometers?) and am aware that indoor environments are a sea of magnetic distortion of every kind. So I do not think a "compass" can help.

My limited understanding is that an IMU (AHRS?) can help correct the heading error, but I am concerned that they may not be able to deal with the normal magnetic disturbances inside of a building. I cannot tell from online research if such IMUs are able to compensate for variable magnetic disturbances, such as are found in a building, or if they will be just as poor at estimating a heading as a digital "compass".

I've found two low cost IMUs: CH Robotic's UM 7: www.chrobotics.com/shop/um7-lt-orient... The 9DOF Razor IMU: www.sparkfun.com/products/10736

Will either of these provide an accurate enough heading estimate to correct error caused by wheel slip during rotation in an indoor environment?

edit retag flag offensive close merge delete

1 Answer

Sort by » oldest newest most voted
3

answered 2014-06-28 08:33:30 -0600

updated 2014-06-30 09:20:11 -0600

Measuring the angular rate around the z-axis is performed using a single gydro on the Turtlebot, see Calibrate Odometry and Gyro. Replicating that setup seems like a straightforward approach. You can of course also use a 6 or 9 DOF IMU for the task and just use the z-axis gyro data, ignoring the other axes. Note that you´ll be able to reduce error due to slip, but noise in sensor data will always results in some odometry drift, so you´ll probably want to use AMCL or gmapping to get proper localization and/or mapping.

edit flag offensive delete link more

Comments

That connected the dots for me. I can get a simple gyro much cheaper and experiment with it using the Turtlebot as an example. It seems obvious now, but I needed a pointer. Thank you.

ChrisL8 gravatar image ChrisL8  ( 2014-06-29 19:33:43 -0600 )edit

Question Tools

2 followers

Stats

Asked: 2014-06-27 10:14:17 -0600

Seen: 1,587 times

Last updated: Jun 30 '14