* @brief Bundled heading state for thread-safe access via DataBuffer. * * Captures both the initial IMU base quaternion (set when heading is * rein

* @brief Per-tick locomotion command passed from input interfaces to the planner. * * Written into a DataBuffer<MovementState> by the input thread a

Clean separation of different data types

PyVista-based visualizer for VR 3-point pose debugging. Coordinate convention: - X: forward (RED axis) - Y: left (GREEN axis) -

A class that creates a reduced order robot model by fixing certain joints. This class maintains a mapping between the reduced state space and

Manages the ROS2 node and executor. Usage example: ```python def main(): ros_manager = ROSManager() node = ros_manag

Support Batch inference

Standardized message schema for camera images. Handles two encodings on the wire: * **str** – legacy base64-encoded JPEG. * **bytes** –

Publishes any serializable dict to a topic.

ZMQ client that deserializes merged camera frames from the server.

This is the error raised by :class:`cv_bridge.CvBridge` methods when they fail.

This dataclass stores additional information that is relevant to the data collection process.

Returns the absolute pose of the iPhone device relative to the world frame. The coordinate system is defined when ARKit is initialized.

This is a standardized message schema for image data. Any camera should use this schema to serialize (send to queue) and deserialize (rec

Context manager for mocking ROS environment and subscribers.

Class for visualizing data using Rerun.

Handles timing and performance monitoring for different code sections.

Places all objects within the table uniformly random. Args: name (str): Name of this sampler. mujoco_objects (None or Mujoc

@brief exception indicating access out of the defined range @sa https://json.nlohmann.me/api/basic_json/out_of_range/

@brief exception indicating access out of the defined range @sa https://json.nlohmann.me/api/basic_json/out_of_range/

Support Batch inference

Class that serves as client for multiple different cameras.

Container for handling images. It is equipped with the necessary methods that one needs for loading and saving images and fetching their data

Monitor joint velocities for G1 robot arms and hands.

A streamer for Nintendo Joy-Con controllers following iPhone streamer pattern. Supports two modes: locomotion and manipulation.

Replay policy for Lerobot dataset, so we can replay the dataset and just use the action from the dataset. Args: parquet_path: Path to

Blender object with support for changing the scaling

Configuration for the OAK camera.

Configuration for the OAK camera.

Sensor for the OAK camera family.

Sensor for the OAK camera family (OAK-D, OAK-1, etc.).

* @brief Configuration for planner initialization */

Subscribes to any topics published by a ROSMsgPublisher.

Output SMPL(not smplx)-Neutral 24 joints (Faster, but cannot output vertices)

Synthetic motion dataset for training without the full motion dataset. Each sample is a random tensor of shape [T, feat_dim] where T is drawn

Robot-agnostic teleop retargeting inverse kinematics code. Focus only on IK processing, ignore commands.

Base simulator class that handles initialization and running of simulations

Generates a four-walled bin container with an open top. Args: name (str): Name of this Bin object bin_size (3-array): (x,y,z)

Coffee machine lid piece (used in Coffee task).

Cup object with optional handle.

Episode state controller for data collection. Manages the state transitions for episode recording: - IDLE: Not recording - RECORDING: Cur

Dummy class just takes out the fingers from the data.

Supplemental information for the G1 robot. Args: waist_location: Where to place waist joints in the joint groups elbow_pose:

Standardized message schema for image data.