Quick Start¶

Rio is built around one idea: separate the hardware description from the control logic. A station declares every node in your setup; a script drives any station without modification.

To see all available components:

uv run rio-list-stations   
uv run rio-list-robots     
uv run rio-list-cameras     
uv run rio-list-interfaces

How it works¶

A station is a Python dataclass in examples/cfg/ that declares every hardware node:

arm / gripper — robot driver and gripper
cameras — named dict of camera streams
teleop — input device (motion controller, leader arm, or keyboard)
recorder — saves demonstrations as .vla files

The teleoperation scripts (teleop_eef, teleop_leader_follower) read the station config, wire up all components, and run the control loop. The scripts are platform-agnostic — swap the station to use different hardware without touching the script.

1. Configure a station¶

Pick the config in examples/cfg/ closest to your hardware and update the hardware-specific fields: robot IP or serial port, camera serial numbers, and recording path.

Note

The example scripts select the station via the STATION environment variable (e.g. STATION=Xarm7EEFStation). This resolves the class name from examples/cfg/__init__.py at startup.

See Station Configuration for the full field reference and how to compose your own station from scratch.

2. Pick a control mode and run¶

Rio has two teleoperation paradigms. Choose the one that matches your teleop device.

EEF / Cartesian control¶

Use this when your input device outputs motion deltas (Spacemouse, Gamepad, Keyboard). The script drives the robot in end-effector space.

STATION=Xarm7EEFStation uv run -m examples.teleop_eef

Switch the input device without editing the config:

STATION=Xarm7EEFStation uv run -m examples.teleop_eef --teleop Gamepad
STATION=Xarm7EEFStation uv run -m examples.teleop_eef --teleop Keyboard

Teleop modes (switchable at runtime via a device button):

Mode	Axes active
`TRANSLATION`	XYZ only (default)
`TRANSLATION_ROTATION`	XYZ + RPY
`TRANSLATION_2D`	XY only
`ROTATION`	RPY only

Leader-follower / joint mirroring¶

Use this when your input device is itself a robot arm that the follower mirrors joint-by-joint (Gello, SO100 leader).

STATION=Xarm7GelloStation uv run -m examples.teleop_leader_follower
STATION=SO100Station      uv run -m examples.teleop_leader_follower

On startup the script checks that the leader joints are aligned with the follower — fix any misaligned joints before confirming.

Example stations at a glance¶

Station class	Config file	Control mode	Script
`Xarm7EEFStation`	`examples/cfg/xarm_eef.py`	EEF (Spacemouse)	`teleop_eef`
`Xarm7GelloStation`	`examples/cfg/xarm_gello.py`	Leader-follower (Gello)	`teleop_leader_follower`
`SO100Station`	`examples/cfg/so100.py`	Leader-follower (SO100)	`teleop_leader_follower`
`BimanualSO100Station`	`examples/cfg/bimanual_so100.py`	Leader-follower bimanual	`teleop_leader_follower`

Adding your own station¶

Copy the closest config in examples/cfg/ and rename the class.
Update hardware fields (IP, ports, serials, calibration paths).
Register it in examples/cfg/__init__.py — add the import and class name to __all__.
Run with the appropriate script:

STATION=MyRobotStation uv run -m examples.teleop_eef             # EEF control
STATION=MyRobotStation uv run -m examples.teleop_leader_follower  # joint mirroring

Common CLI overrides¶

Any config field can be overridden on the command line without editing the file:

STATION=Xarm7EEFStation uv run -m examples.teleop_eef \
    --instruction "pick up the can" \
    --visualizer Rerun \
    --freq 30

Recordings are saved as .vla files to recorder_cfg.path. See Collect Demonstrations for replay, conversion, and next steps.