Units and wrapping of joint_pos_rel, joint_vel_rel, base_ang_vel, and root_euler_angle?

[lde-RL]

I’m currently working on a Sim2Real project using an ONNX policy trained in Isaac Lab. To apply the policy on my real robot, I need to align the observations from my code with the real hardware: specifically, Dynamixel MX-64 motors and an IMU sensor.

In Isaac Lab, the observation configuration is as follows:

joint_pos = ObsTerm(
    func=mdp.joint_pos_rel,
    noise=Unoise(n_min=-0.04, n_max=0.04),
    params={
        "asset_cfg": SceneEntityCfg("robot", joint_names=[".*"])
    },
)
joint_vel = ObsTerm(func=mdp.joint_vel_rel, noise=Unoise(n_min=-1.5, n_max=1.5))
base_ang_vel = ObsTerm(func=mdp.base_ang_vel, noise=Unoise(n_min=-0.15, n_max=0.15))
base_euler = ObsTerm(func=mdp.root_euler_angle, noise=Unoise(n_min=-0.125, n_max=0.125))

To ensure compatibility between the simulation and the real robot, I’d like to confirm:

1. What are the units used for joint_pos_rel, joint_vel_rel, base_ang_vel, and root_euler_angle in the ONNX file? Are all angular quantities
   in radians?

2. How is angle wrapping handled for joint_pos_rel and root_euler_angle? For example, do angles wrap within [-π, π], [-2π, 2π], or are
   they unbounded?

3. When mapping real sensor data (from the MX-64 motors and IMU) to match the Isaac Lab observations, do I need to wrap or
   normalize the values in any specific way?

Any clarification would be greatly appreciated!

[RandomOakForest]

Thank you for posting this. We definitely need to have a glossary of units. Angular quantities are radians. Standard radians interval is [0, 2π]. Real sensor data will depend on your sensor input. Let us know if we can be of further help.