（Sim) Very strange movement for /omni_base_controller/command_velocity topic

[y-masutani]

HSR in Gazebo for /omni_base_controller/command_velocity topic is very strange.

To clarify the issue, after launching the simulation, I ran the following command on another terminal in case of linear.x = 1.0.

ros2 topic pub -r 10 -t 50 /omni_base_controller/command_velocity geometry_msgs/msg/Twist "{linear: {x: 1.0, y: 0.0, z: 0.0}, angular: {x: 0.0, y: 0.0 , z: 0.0}}"

Result:

• linear.x = 1.0 : moves forward (normal?)
• linear.x = -1.0: moves backward and then turn
• linear.y = 1.0 : turns left backward 90 degrees, then moves sideways
• angular.z = 1.0: No movement at all

See screencast for linear.y = 1.0.
https://drive.google.com/file/d/1-8tSIVygn3Aq7K3vI5kEcAj2Co07CGmu/view?usp=sharing

How can I solve this problem?
Has anyone else encountered this problem?

[y-masutani]

After that, the action communication /omni_base_controller/follow_joint_trajectory was also investigated, with similar results. Moreover, I confirmed that odometry /omni_base_controller/wheel_odom was at the expected value.

See the windows of Gazebo and RViz (odometry) during the action communication.
https://drive.google.com/file/d/1__IjnBgEoyqWRvjeIb3HsUSPJcFwTESI/view?usp=sharing

[y-masutani]

I found that the wheels and the floor are slipping in Gazebo.

When Twist.angular.z command is given, base_roll_link rotates with respect to base_link, but base_link does not rotate at all with respect to the floor because the wheels and the floor are slipping.

Watch the screencast to see how it works. It is displayed in wireframe for clarity.

Screencast

Does this phenomenon reproduce in your environment?
How can we solve this problem?

[y-masutani]

https://github.com/orgs/HumanSupportRobot/discussions/97