search

google-deepmind / mujoco_mpc

Real-time behaviour synthesis with MuJoCo, using Predictive Control
https://github.com/deepmind/mujoco_mpc
Apache License 2.0
981 stars 148 forks source link

Questions regarding using stateful actuator intvelocity #313

Closed MiaoDragon closed 6 months ago

MiaoDragon commented 6 months ago

Hi,

I'm a PhD student trying to use this package as a library for controlling task in mujoco. First thanks a lot for this package. I have successfully integrated it and be able to solve example tasks in mujoco. However, I have some questions regarding using this with stateful actuator, such as intvelocity. What I'm trying to do is to use intvelocity for tracking the position and velocity input. However, I found that mujoco-mpc does not treat data->act as part of the control, but rather as a state. Hence, I'm not able to sample the position control in data->act for my task. I'd like to ask your input for my issue. And could you recommend any way to control the position and velocity at the same time for MPC task? I found that intvelocity performs better than a combination of position and velocity actuators, since I assume it's the correct way of implementing PID controller.

Thanks. I would really appreciate any help!

(I also opened another issue under mujoco at issue)

thowell commented 6 months ago

Since Yuval has responded on the MuJoCo thread I will close this issue.