about control frequency

[tianhaowuhz]

Hi, Im recently trying to use this api, and Im wondering what's the highest control frequency can this api reach, I have tested like moving continuously, and it's about 1s for each action, is there is a way to have higher frequency?

[justagist]

Hi, what actions do you mean? If you are using the low-level controllers, they should be able to operate at a much higher frequency. I think I have been able to send commands from python at the rate of over 250 commands per second, while the actual low-level control loop in C++ (tries to) run at 1kHz. It doesn't actually reach 1kHz in practice and is closer to 500 - 700Hz in my observations.

[tianhaowuhz]

I am so sorry, there seems no notification, i didn't notice you already respond to me. I am trying to use exec cmd in python wrapper, is these the command you mentioned that can reach the rate of 250 commands per second?

[tianhaowuhz]

also ,I am wondering which api in python should I use if I want to make panda move more smooth like when track a marker?

[justagist]

I am so sorry, there seems no notification, i didn't notice you already respond to me. I am trying to use exec cmd in python wrapper, is these the command you mentioned that can reach the rate of 250 commands per second?

Yes, I am able to attain over 250Hz with the python commands. Note that this requires you to send commands in a loop that is running independently at at least that rate. Simple example here.

also ,I am wondering which api in python should I use if I want to make panda move more smooth like when track a marker?

You can still use the python code to test provide smooth motion; you have to make sure that the provided trajectory is very smooth and continuous, especially if you use position control (see #28). But motion should be much smoother if you use torque or velocity control. If you want the robot tracking a marker, I would suggest writing a task-space position/impedance controller using the torque command interface (a vaguely similar example is the task-space control demo in panda_simulator).