PID tuning for a closed-loop configuration robot?

edit

Hi,

I am fairly new to PID tuning in Gazebo. I am working on a project that involves a closed-loop configuration robot (where the end effector is attached back to the base). Part of this project requires the tuning of the Joint Position Controllers that each control some revolute joint. In the past two months, I've noticed that most PID tuning videos in Gazebo involve a robot arm or some open-loop configuration. You would start with tuning the base and continue in sequence to the other links.

However, I am unsure if it's due to our inexperience, but my team and I have been having a significantly hard time trying to tune the Joint Position Controllers in our closed-loop configuration robot. We have been making use of the position plot function in Gazebo and the ROS rqt-gui. The inaccuracies of the PID gains are most apparent when some joint experiences a spike in velocity and the whole robot goes flying (because it's not anchored to the world). While the fix would be to anchor the robot, we are experimenting with the robot's locomotion capabilities and cannot simply just anchor it

Tldr; does anyone have any advice for PID tuning for closed-loop configuration robots? Is it the same as PID tuning for an open-loop configuration like a robot arm where the end effector is NOT attached to the base?

Thanks in advance