openai / mujoco-py

MuJoCo is a physics engine for detailed, efficient rigid body simulations with contacts. mujoco-py allows using MuJoCo from Python 3.
Other
2.87k stars 813 forks source link

How to get document of mujoco-py #414

Open 1576012404 opened 5 years ago

1576012404 commented 5 years ago

i want to get document of usage of API of mujoco-py for example,the meaning of model.data.com_subtree model.data.qpos model.data.qvel

MillionIntegrals commented 5 years ago

Most of these things are very thing wrappers on top of underlying C structures from mujoco itself. Documentation is here. http://mujoco.org/book/APIreference.html

wxplovehlt commented 5 years ago

Most of these things are very thing wrappers on top of underlying C structures from mujoco itself. Documentation is here. http://mujoco.org/book/APIreference.html

what does mean"To include mujoco-py in your own package",where mujoco-py should be located?

tuonan commented 5 years ago

@MillionIntegrals Thanks . Can you understand sim.data.qpos ,sim.data.qvel?I watch it,butBut its explanation is too little and too vague for me.

MillionIntegrals commented 5 years ago

qpos is basically a position of each joint in the simulation and qvel are the velocities of these joints.

wxplovehlt commented 5 years ago

what about the action[3] in the env Fetchpickandplace.the value of it means?

robotlearning123 commented 5 years ago

The action space of Fetch environment is 4. Specifically, action[:3] represents the position of end-effector: (x,y,z); action[3] is the gripper control with one DOF. From the xml file, we can know that the gripper has two actuators, so in the FetchEnv, it uses one binary signal to control two actuators. We can get a similar use from StanfordVL-robosuite.

robotlearning123 commented 5 years ago

Except the MuJoCo officially document, I also try to get the meaning and explanation from mujoco-py source code. You can just search the data type or function name in the source code and almost all have the explanation in the C++ code.