.. _custom_sim:

Customize Env 
===============================

.. code-block:: bash 
    
    python3 tele.py --env <env_config> --tele <tele_config> 

The teleop script takes in two configurations: one for environment and one for teleoperation. 

The teleoperation config is tightly coupled with robot configurations -- each robot has its own teleoperation config.

To use **the same robot across different environments**, you simply need to create a new environment in ManiSkill3 and modify the **environment config** accordingly.

.. dropdown:: Why this design?
    :open:

    It decouples robot control logic from the task/environment logic, so that in real-world teleoperation, the same robot and control interface can be reused consistently, while only the environment-specific settings need to be changed.

Environment Config
---------------------------

.. code-block:: yaml

    # these definitions follow ManiSkill3 definition.
    env_name: str = "PickCube-v1" # name of the environment 
    robot_uids: str | None = None # name of the robot

    ignore_terminations: bool = True  
    # ignore the terminations for all mani_skill envs, you can skip it. 

    obs_mode: str = "state" 
    action_mode: str | None = None
    render_mode: str = "rgb_array"  
    sim_backend: str = "cpu"

    control_mode: str = "pd_pos_control"
    sim_config: NewSimConfig = field(default_factory=NewSimConfig)

    # these are teleoperation related definations. 
    root_pose: PoseConfig = PoseConfig() 
    # Root pose of VR display - changes your VR view
    goal_site: PoseConfig = PoseConfig(p=[0.0, 0.0, 1.0]) 
    # Pose of green sphere for recording status visualization
    # If the environment has a goal site defined in ManiSkill3 Env (e.g. goal position for pick up task),
    # it will use that goal site pose instead of this one
    
    success_condition: bool = True 
    # whether have success condition or not
    auto_record_success: bool = False 
    # If True, recording stops automatically when success condition is met
    auto_record_success_threshold: int = 20 
    # Control steps to wait before stopping recording after success
    # For example, at 10 Hz control frequency, recording stops 2 seconds after success

You can preview the environment by provided tools::

    python3 tools view <env_config>

See example configs in `vr_teleop/configs/examples <https://github.com/liuyulinn/imvr/tree/master/vr_teleop/configs/examples>`_.