Get the current simulator state through the handler state-mode contract.

Decode raw bytes into axis and button state.

Add the observation to the list.

Set the seed for the environment. Args: seed: The seed for random generator. Defaults to -1. Returns: The se

Save the images or videos.

Step through all environments

:return: The current size of the buffer

Set simulator state for selected env indices.

Scan ``data.contact[:ncon]`` and integrate per-primary readings.

Close the environment.

Return EE state using site queries. Returns: ee_pos_world: (B, 3) gripper position from site ee_mat_world: (B, 9) gri

World position (3,) of object ``name``.

Render the environment.

Add a new batch of transitions to the buffer

Advance one control step using the current policy.

:param batch_size: Number of element to sample :return:

Return the canonical native-asset paths as strings.

Update episode tracking with new rewards and done flags.

Close the environment.

Recursively apply functions to a nested dictionary or list or tuple, given a dictionary of {data_type: function_to_apply}. Args:

Execute actions and return observations, rewards, dones, extras. RSL-RL expects combined done flags (terminated OR truncated).

Get the current simulator state through the handler state-mode contract.

Learn input values using Welford's online algorithm

Set arm goal + advance gripper target one control step. Call per control step.

Resets the specified environments and returns observations. This function calls the :meth:`_reset_idx` function to reset the specified enviro

The device on which the environment is running.

Reset the buffer.

Get a numpy array with access to the shared memory

Look up a task by either its robosuite env name or its metasim name.

Reset the environment.

Return slice(i, i+n) and advance internal counter.

Build a native robosuite env for ``task`` with a fixed, deterministic config.

Render the environment.

Called before each simulation step in one control step.

Remember the site-id once the handler is known.

Compute the actuator model.

Return {"arm": ..., "gripper": ...} for the deployed control mode (vendored).

Load a visual-matching overlay exactly as SimplerEnv does: BGR->RGB, /255 float.

Quaternion [x,y,z,w] -> 3x3 rotation matrix (robosuite convention).

Register all SimplerEnv tasks under the ``SimplerEnv/<task>`` namespace. The task list is taken from ``simpler_env.ENVIRONMENTS`` (the authoritat

World position of the panda ``panda_hand_tcp`` link.

Convenience method to convert pymunk coordinates to pygame surface local coordinates. Note that in case positive_y_is_up is False, this funct

Learn input values using Welford's online algorithm

Unique key for ``name``, suffixing duplicates as ``name#1``, ``name#2``, ... RoboTwin tasks may create several actors with the SAME name (e.g.

Return initial states for all envs. Format: list[dict] of length num_envs Each dict contains: "objects", "robots", "cameras", "extras