search

isaac-sim / IsaacLab

Unified framework for robot learning built on NVIDIA Isaac Sim
https://isaac-sim.github.io/IsaacLab
Other
1.91k stars 735 forks source link

Command and Action #337

Closed Nimingez closed 2 months ago

Nimingez commented 5 months ago

Thank you for you great work！ Robot reach task， reach_env_cfg CommadCfg UniformPoseCommand and ActionCfg I noticed that your command provides the end-effector position, while the policy outputs actions corresponding to joint positions. Where does the process of inverse kinematics take place?

zoctipus commented 5 months ago

You can take a look at 'source/extensions/omni.isaac.orbit/omni/isaac/orbit/controllers/differential_ik.py' method: _compute_delta_joint_pos

The implementation how ik is calculated is there

Mayankm96 commented 5 months ago

Exactly. The action term we are using there is DifferentialInverseKinematicsActionCfg. It converts pose absolute or relative commands to joint commands and applies them into the simulation.