Summary

This repository implements a multi purpose controller for manipulators using the ros2-control framework and Gazebo for simulation. The controller exploits KDL and urdf parsing to resize its variables accordingly to the DOFs of the manipulator and implements the following control techniques:

• PD with gravity compensation
• Computed torque
• Backstepping
• Adaptive Backstepping

The robots_description package implements a modified franka panda URDF description where the joints have been modified to be without limits, while controller_error_msgs contains the messages used to monitor the controller error, send the desired pose and twist, and set desired gains.

Note on the Adaptive Backstepping Controller

For the adaptive controllers, a regressor formulation is needed, so it requires a symbolic formulation of the robot. The files panda_gen.cpp, thunder_panda.cpp and all the headers int he panda_generated folder are, as the name suggests, generated by the thunder_dinamycs library.

Using `ros2 run manipulator_controller send_pose.py` will launch a GUI that allows to send the desired pose to the controller. The "Toggle Cricle" button will send the controller the desired poses and twists to move in a circle on the selected plane, with the selected radius and centered in the desired point.\ The last 4 spinboxes allow to change the controller gains during runtime.

Usage

To launch the Gazebo simulation, after compiling and sourcing the environment, run ros2 launch robots_description panda_gazebo.launch.py. This will launch both the controller (and the robot_state_publisher) and the simulation.

Configuration

The configuration of the controller can be done by modifying panda_multi_controller.yaml file inside the robots_description package. The parameters are the following:

• joints: list of all the joints offering the selected command/state interfaces
• description_topic: the name of the topic where the robot's description is being published
• base_link: name of the link acting as the base of the manipulator
• ee_link: name of the link considered the end effector; this can be any link along the cinematic chain (different from base_link)
• controller_type: selects type of control to use:
  • PD - PD with Gravity compensation
  • CT - Computed Torque
  • B - Backstepping
  • A - Adaptive Backstepping
• Kp_p/Kv_p: position gains for proportional/derivative gain matrices
• Kp_o/Kv_o: orientation gains for proportional/derivative gain matrices
• Kv: only present if the derivative gain matrix is in joint space;
• r: (Adaptive only) regressor gain