Skip to content

Getting Started

Back to Star Arm 102

Use this page to choose the right path and complete your first connection and control checks. Preassembled arms can skip DIY assembly. You do not need every development framework for first control.

1. Identify your equipment and goal

Model or setup Purpose Next step
102-LD Unpowered leader that captures the operator's pose Connect the leader and check joint feedback or virtual following
102-HD Powered leader; button behavior depends on hardware and software Check the supplied configuration and select a path that supports HD
102-FL Follower that executes movements Check communication and calibration; it can run a matching pretrained policy independently
LD / HD + FL Teleoperation within the same series Start with the web tool or the applicable Python example
102 leader + reBot Teleoperation across products Follow the reBot integration guide

Confirm the model from its label, order, and delivery documentation. See Specifications for hardware values; do not apply LD power or servo configurations to other models.

2. Prepare your workspace and equipment

  • Secure the arm base and clear cables and objects from joint travel paths.
  • Prepare the matching power supply, controller, and USB data cable. Follow the supplied wiring instructions.
  • Identify the leader and follower USB connections and record their actual ports.
  • Basic communication and teleoperation do not require cameras. The ACT example needs two cameras and a matching scene.
  • DIY users should review the assembly guide and check parts, servo locations, and wiring first.

Success check: the model and power supply match, the base is secure, cables have enough slack, and you can stop control and disconnect power.

3. Check communication

Open the Robot Arm Web Control and Configuration Guide and select the model and port as instructed. Your browser must support the required serial port features; use the environments listed on the tool page.

  1. Close other applications using the serial port.
  2. Connect the device being checked and confirm its model and port.
  3. Check connection status and joint feedback before connecting the second arm.
  4. With a leader, use virtual FL mode first to check pose capture and mapping.

Success check: the device connects and joint readings follow its physical pose. Leader and follower ports are distinct. A sent command does not confirm that the real arm has reached its target.

Important

Restoring factory parameters, resetting hardware zero, and LeRobot calibration are different operations. Do not rewrite parameters or zero positions just to check communication. When needed, follow the official procedure for your model.

4. Complete first control

For a leader/follower set, follow the web guide to connect the real FL, check direction with slow, small leader movements, then stop and disconnect. For FL only, preview a small joint target in the tool and confirm it before execution. Stop on unexpected motion or direction and use Troubleshooting.

For code-based control, open the Python SDK guide and check the example's supported model and actual serial port. For LeRobot, complete its calibration steps first and keep the same device ID in later commands.

Success check: small movements have the expected direction, feedback matches the real arm, and stopping works. Do not run an autonomous policy before passing this check.

5. Versions and compatibility

Path Purpose Environment and version reference
Web tool Connection, pose feedback, teleoperation, trajectory configuration Use the versions and browser requirements on the tool page
Python SDK Basic control and teleoperation Check the model against the Python guide and repository example
Released ACT example FL + two cameras running a supplied policy The release guide uses Ubuntu 22.04, Python 3.10, LeRobot 0.4.1, and both StarArm plugins at 0.0.1
Other LeRobot integrations Teleoperation, data collection, and training Check models, device types, and dependencies in the plugin directory first
ROS 2 MoveIt planning, Gazebo simulation, and system integration The current ROS 2 guide uses Ubuntu 22.04 and ROS 2 Humble

These entries describe the references for each path, not newly completed hardware validation of every combination. Python and web control do not require ROS 2. Check GPU requirements for the specific training or inference task.

The ACT release uses the lerobot_robot_stararm102 device type. Types such as stararm102_fl from other integrations are not direct replacements: joint names, units, and calibration must match the model. Keep different plugin paths in separate environments to avoid replacing packages with the same name or mixing calibration files.

Before upgrading, record the operating system, plugin and code versions, device IDs, and camera names and models. Use the version-pinned run guide for the ACT installation and execution commands.

Next steps

Try the Pretrained ACT Model Learn Your Own Task

For CAD, printing files, or assembly resources, open Hardware & Open Resources.