Is there any specific decision point or difficulties when tuning Kp / Kd to Simulating or Sim2Real Asimov?
i think the main thing we realised for kp kd tuning for sim and real is that its treated as two diff systems to tune.
what i mean by this is what we found out was that simulation any kp kd usually causes the robot to walk well or achieve the objectives but it has to be in certain ratio ie not overdamp/underdamped so it doesnt learn any oscillations in sim which become worse and aggravate in real. but for real what we realised is its mainly a tracking problem so as long as your real kp kd is able to track your policy’s output is fine so actually you can tune them separately or the best way to do it is find the kp kd in real first that you know will work or provide enough torque so your system isnt loose and then use those for sim since sim isnt very sensitive to kp kd values just the ratio of them.