[Example 11] Carlike Robot using Bicycle Steering Controller

[ARK3r]

Example with 2 front wheels and 2 rear wheels.

Copying a lot from DiffBot, I would say the only difference is that CarlikeBot uses Ackermann Steering Controller, everything else should be quite similar.

I've also added Carlikebot to ros2_control_demo_description as per the new convention for the robot models.

[ARK3r]

Here I changed the carlikebot description a little so that I can use the bicycle controller on a 4-wheel vehicle. There are two virtual joints, one for steering and one for traction where carlikebot's actual wheels just mimic the movements of the virtual joints. I think this will be a good baseline for using ackermann on the same urdf with minor changes and adding a couple of arguments to the xacro and launch files.

I have yet to fix the documentation but everything should be functional so I would appreciate it if you could give it a try. I also made some design changes by defining a map for hw_interfaces_ instead of a vector. I believe this will make the code way more readable.

Also I added a node to relay the /tf_odometry topic to /tf which you can activate by calling ros2 launch ros2_control_demo_example_11 carlikebot.launch.py relay_odometry_tf:=true

[ARK3r]

One strange thing I'm noticing is when linear.x=0, whatever value angular.z gets, the steering just snaps to $\pm \pi$ which I believe comes from SteeringOdodmetry::get_commands. I'm just not sure what the reason for this is and it feels strange. Wonder if anyone has an explanation.

[ARK3r]

There is an issue in the URDF: the traction joint of the front wheels has wrong axis.

this issue is now resolved with the new description.

[christophfroehlich]

One strange thing I'm noticing is when linear.x=0, whatever value angular.z gets, the steering just snaps to ±π which I believe comes from SteeringOdodmetry::get_commands. I'm just not sure what the reason for this is and it feels strange. Wonder if anyone has an explanation.

I think this is fine, because this is a singular point of the inverse kinematics. It would be also fine if the bicycle controller does not accept it, because this does not work with the traction wheel not being the steered one (but with the tricycle model it would work).

There is an issue in the URDF: the traction joint of the front wheels has wrong axis.

this issue is now resolved with the new description.

Perfect.

Here I changed the carlikebot description a little so that I can use the bicycle controller on a 4-wheel vehicle. There are two virtual joints, one for steering and one for traction where carlikebot's actual wheels just mimic the movements of the virtual joints. I think this will be a good baseline for using ackermann on the same urdf with minor changes and adding a couple of arguments to the xacro and launch files.

I'm fine with that. This also is aligned with my nomenclature of ros-controls/ros2_controllers#954 (I'd be happy about some feedback there).

Also I added a node to relay the /tf_odometry topic to /tf which you can activate by calling ros2 launch ros2_control_demo_example_11 carlikebot.launch.py relay_odometry_tf:=true

I changed this to use remapping instead, to avoid a dependency on topic_tools.

[christophfroehlich]

@Mergifyio backport iron humble

[mergify]

backport iron humble

✅ Backports have been created

• #449 [Example 11] Carlike Robot using Bicycle Steering Controller (backport #316) has been created for branch iron
• #450 [Example 11] Carlike Robot using Bicycle Steering Controller (backport #316) has been created for branch humble