ROS 2 Robotiq

[PeterMitrano]

Tasks:

• [x] read the manual
• [x] modify the interface of driver.hpp to support the 3f
• [x] modify default_driver.hpp/.cpp to match
• [x] modify the hardware interface to match
• [x] make a launch file to visualize the robot description
• [x] Try to make transmissions that map 4 actuators to the full kinematic state of all the joints (individual mode)
• [x] make the launch file start the right controllers ~- [ ] Write a custom controller to offer a clean state/cmd API~
• [x] consider rewriting the launch file in XML
• [ ] Write a node that lets us control the grippers from an xbox gamepad
• [ ] Merge into PickNik repo
• [ ] Make the hardware tests work so CI passes
• [ ] Figure out how MoveIt can integrate with my GripperCommand action

[PeterMitrano]

One issue I'm having is that we can only export double datatypes from the hardware interface (for either joint or GPIO types) which means our enum type for the discrete modes or object detection data can't be represented easily. I could conceivably convert to/from doubles, as is done in the epick driver but that seems very ugly.

[PeterMitrano]

For the custom controller we probably want an API like this:

• ROS service for changing the mode, since this is actually quite a long/slow process. So you probably want it to block until complete.
• ROS topic for simple command mode (custom msg)
• ROS topic for individual command mode (custom msg)
• Publish all the state not already being published by the JointStateBroadcaster, which includes things like the actuator states, the object detection states, and the actautor currents. Honestly it feels like ROS2 control is making me write lots of bad boilerplate when the only standard component I'm using in the end is the JointStateBroadcaster, and even then I'm having to tell it every little thing abotu how to use my robot. I wonder if I'd be better off just writing a standalone C++ ROS Node and ditching ROS2 control...