ROS 2 integration

ROS 2 can coordinate perception, planning, and robot control, but the compute application should not assume that a middleware message is safe to apply directly to a joint drive.

Define the boundary

Keep a small control adapter between ROS 2 topics and the robot's actuator protocol. Validate command range, timing, stale-message behavior, and enable state inside that adapter. Give the adapter a tested stop path independent of perception or UI processes.

Bring up in stages

1. Run the compute image and verify network access without motor power.
2. Start the ROS 2 graph and inspect the expected nodes and topics.
3. Replay recorded sensor data through the perception pipeline.
4. Connect the control adapter to a simulator or unpowered hardware interface.
5. Only after electrical and mechanical checks, enable actuators under supervision.

The exact ROS 2 distribution and installed packages depend on the image. Check them locally before using a package command copied from another release.

The Qmini project guides describe its published model and control resources; those resources still require validation against the physical build in front of you.