Robot moves only forward for Twist msgs with Diff drive plugin

[ashBabu]

Ubuntu 22.04, ROS2 Humble, Gazebo Fortress. Installed everything from apt repos

I have added a diff_drive plugin for my robot as shown in here. Then I launch the simulation using this launch file and press play button (this is done after commenting out the odometry and velocity plugins in the urdf). After which I tried both the following separately.

• ign topic -t "/cmd_vel" -m ignition.msgs.Twist -p "linear: {x: 1.0}, angular: {z: 1.0}"

• ros2 topic pub /cmd_vel geometry_msgs/msg/Twist "{linear: {x: 1.0, y: 0.0, z: 0.0}, angular: {x: 0.0, y: 0.0, z: 1.0}}"

Even though I have specified the angular velocity component, the robot moves only straight as shown in the video. I am also expecting an odom --> base_link transform to be published but is missing.

All the work that I am doing is available in my repo skidy and the README provides additional information

[scpeters]

I'm wondering if this is an issue with the physics and wheel surface parameters or if it is a problem elsewhere. To debug this, can you try lifting the wheels off the ground and plotting the joint states after sending a /cmd_vel? If the wheels spin at different rates when it's off the ground, then it may be a physics issue.

To lift the wheels off the ground, you can try placing the robot body on a box so that the wheels can spin freely, or by setting gravity to 0 and raising the vehicle off the ground

[azeey]

Looking at your robot_diff_drive.xacro file, I don't see any friction parameters specified. You can use parameters from the diff_drive example in gz-sim as a starting point: https://github.com/gazebosim/gz-sim/blob/9d79b629c9e0fed1c8635e30f06b2c61462f32d8/examples/worlds/diff_drive.sdf#L149-L155

[ashBabu]

@scpeters : I have tested with gravity = 0, raising the vehicle above ground. The resulting video is here. A rosbag of the /joint_states is available here and seems empty. @azeey : I have added the friction parameters and tested but the result looked exactly the same as before. Out of curiosity, I send a linear only Twist msg to the robot when gravity=0 and robot at 2.5m above ground and it still moved forward video. If friction with ground is necessary for rotation of wheel, this should not happen rt?

[ashBabu]

I think gravity has no effect. Even though I have set <gravity>0 0 -9.8</gravity> in the world file, the robot doesn't fall to the ground quickly when I set initial position.z = 1. It's falling down so slow.

https://github.com/gazebosim/gz-sim/assets/10324110/775f71ea-09fe-439a-9b0b-9c609cb9b36f

[scpeters]

can you try to see if the wheels on the left and right side of the vehicle are spinning at different rates? For a diff-drive, when you command a twist with a non-zero angular z component, there should be a difference in the left and right wheel spin rates

[scpeters]

can you try to see if the wheels on the left and right side of the vehicle are spinning at different rates? For a diff-drive, when you command a twist with a non-zero angular z component, there should be a difference in the left and right wheel spin rates

have you had a chance to try my last suggestion?

[ashBabu]

Apologies for the delay. The wheels seem to rotate at different velocities.

https://github.com/gazebosim/gz-sim/assets/10324110/40142514-94ff-4e0c-b80a-c569a18bfb59

https://github.com/gazebosim/gz-sim/assets/10324110/477dae6b-530e-45ff-bb66-2dd47f80a4f7

1. odom --> base_link from diff drive plugin not working
2. I think because of the above, the /tf tree has only base_link --> drivewhl_l/r_joint.

I have setup the ros-gz-bridge as shown here.

[scpeters]

ok, it does look like your wheels are spinning properly. Looking at your xacro file once more, I see that your joint axes are 0 1 0, so can you try changing the fdir1 parameter from 0 0 1 to 0 1 0? Those should match

[ashBabu]

Thanks @scpeters : It works now. Closing this

[dylan-gonzalez]

I have the same problem. I have the same friction parameters (except my joint axes are "0 0 1"), and I have made sure the fdir1 parameter matches this.

The only difference is that I am using a URDF with <gazebo> tags like so:

  <gazebo reference="left_wheel">
    <surface>
      <friction>
        <ode>
          <mu value="1.0"/>
          <mu2 value="1.0"/>
          <slip1 value="0.035" />
          <slip2 value="0" />
          <kp value="10000000.0"/>
          <kd value="1.0"/>
          <fdir1 value="0 0 1"/>
          <material>Gazebo/Grey</material>
          <turnGravityOff>false</turnGravityOff>
        </ode>
      </friction>
    </surface>
  </gazebo>

Edit: I realise you were also using a URDF file

Update: I fixed my issue as my joint limits (velocity + effort) were too low.