Now that you've gotten comfortable with linux, Dynamixel, ROS, rviz, etc., the next step is to actively move the robot arm. This involves controlling the manipulator in joint-space and in Cartesian-space.

Joint space control is done by identifying the proper ROS topic to publish to within the manipulator interface. It might not be the same one as for the raw Dynamixel, but may be an intermediate one that then publishes to the raw Dynamixel topics.

Cartesian space control definitely works differently. This requires some wrapper libraries or APIs for managing the inverse kinematics that perform the translation from Cartesian space to joint space. One such API is _MoveIt!_, which has been around for over a decade now. It is reasonable functional. Demonstration involves planning a trajectory and executing it through the _MoveIt!_ interface of Rviz or through a C++ interface. (Not sure if there is a python interface, or it might be buggy)

Only one arm needs to be controlled as bi-manual operation will not be the focus of the project thread in this Module.

The lab will be composed of two parts. The first part will be controlling each joint of the manipulator and then plan a trajectory and execute it in the second part.

For part 1, you can refer to this tutorial. The whole point is to understand how rostopic works and how to control the motor via rostopic.

For part 2, this tutorial is a good starting point. It is about the commercial robot arm Franka but can be transferred to Edy. The most related tutorial will be MoveIt Quickstart in RViz, which is the second one. You should understand how to plan a trajectory and let the arm to perform the execution. For this week, you can either achieve this via RViz interface or C++ interface, which is introduced in the third tutorial.

About delivery, for part 1, you should be able to control each motor to rotate to certain degrees. For part 2, you should demo to me that the manipulator performs any valid trajectories.

Back – Main