IMU simulator from hector_gazebo_plugins is noisy even when configured to be pristine

asked 2017-08-16 18:22:56 -0500

Galto2000 gravatar image

Howdy folks,

I am using GazeboRosImu from hector_gazebo_plugins ( ) and I noticed that even when I set my biases, drifts, offsets, etc.. all to zero, I still get noisy accelerations passed through.

When digging a little bit into the code (gazebo_ros_imu.cpp) I noticed that the call to Link::GetWorldLinearVel() is noisy and has a bias. Since this linear velocity is differentiated in gazebo_ros_imu.cpp to produce the linear acceleration, it will also be noisy as a result.

Is this normal Gazebo behavior?



edit retag flag offensive close merge delete


Have you got somewhere with this? I have the same problem. I even tried to calculate linear velocity based on the position/dt - still the result is noisy.

Ago gravatar image Ago  ( 2017-09-26 03:16:25 -0500 )edit

Not really, but I am planning on posting this question in the Gazebo forum shortly. If I get somewhere with this I'll surely post it here.

Galto2000 gravatar image Galto2000  ( 2017-09-26 08:08:00 -0500 )edit

Did you figure out the problem? I seem to have the same issue. I'm simulating the sensor on a Dataspeed car.

Salahuddin_Khan gravatar image Salahuddin_Khan  ( 2017-12-05 01:14:23 -0500 )edit

I posted the question in Gazebo Answers but I haven't gotten any replies. I suspect it might be oscillations in the robot-ground collision dynamics. I haven't had much time to dig deeper into it, but if I had, I'd start with reproducing the issue using a simple cube.

Galto2000 gravatar image Galto2000  ( 2017-12-05 07:34:09 -0500 )edit

This problem still persists. I have found that a simple rolling cylinder does not produce this noise, however a simple, four-wheeled cart does. The noise reduces when the mass of the cart body increases with respect to the wheels. Still no idea how to solve it though

Vianne gravatar image Vianne  ( 2020-07-20 08:24:27 -0500 )edit