URDF Introduction: ROS2 GoPiGo3 gpgMin.urdf

GoPiGo
21

Is it necessary to have a representation of the LIDAR, accurate to fractions of a millimeter, or is the fact that a LIDAR exists at point “X, Y” sufficient?

3 Likes
22

The visual representation is used only by the visualization package, so no it does not have to be accurate in size or even shape.

What is really important is that the x,y,z and Euler rotations r, p, y are as accurate as possible.

http://wiki.ros.org/urdf/XML/link

2 Likes
23

The shape and mass are used for simulation as well. If you’re wanting to simulate your robot in Gazebo, you’ll want something approximately correct. But certainly doesn’t have to be down to fraction of a millimeter - even a simple box or cylinder would be fine. @cyclicalobsessive is correct is that having the location be as accurate as possible is the most important thing (both IRL and in Gazebo).
/K

2 Likes
24

Can Gazebo use a URDF or does it have to have an SDF?

I did some Gazebo simulation with a Create3 SDF, but didn’t ever investigate sim with the GoPiGo3.

Did you create a basic GoPiGo3 SDF for simulation? (I’m afraid of all that xacro stuff most folks are using - it looks good and it adds accuracy, but as you know I like to play with my bot on the floor most.)

1 Like
25

You’re right - it really needs an SDF. I was conflating the two. The SDF is what drives the simulation - but I’ve always built my SDF from the URDF.
/K

2 Likes
26

What happens if Dave moves forward 2 meters, turns around, and moved back to appx. 0.5 meters?.

3 Likes
27

@beetecaste I did not perform that experiment (2m fwd, 180 turn, 1.5m “fwd”) but I can tell you what the result would be based on my other tests:

  • estimated heading accuracy is poor (roughly 5%) resulting in odom estimates diverging faster than actual
  • actual heading accuracy while driving straight is roughly 1% (wheel speeds are fed to bot’s PID on platform) which results in +/- 2-3cm physical y deviation in 2m
  • actual turn accuracy is roughly +/- 1 degree which can add to odom estimate inaccuracy

I gave up trying to integrate IMU with an EKF - I read in one paper that I should not have used the Euler fusion processor output, but in general the EKF configuration was beyond me.

2 Likes