URML (open robot intent language): a validated intent layer above blue (BlueROV2) — request for comment

[idoco2003]

Hi Robotic Decision Making Lab,

URML (urml.dev) is a small, Apache-2.0 language for describing robot intent: a person writes an English sentence, URML turns it into a typed primitive, validates it against the robot's declared capabilities and a safety envelope, then dispatches to whatever runs below. blue is the closest external analog to URML's own marine-runtime, which already drives a BlueROV2 over ArduSub/MAVLink — so I am writing because we are building toward the same substrate from two directions.

Nothing here asks the lab to adopt, host, or maintain anything. This is a request for comment, and genuinely a design question.

URML's marine-runtime targets the same BlueROV2 / ArduSub pairing blue orchestrates, and its ROS 2 runtime meets blue on the ROS 2 side. Your auv_controllers (AUV/UVMS on ros2_control) is the control seam URML dispatches validated motion intent to — the same ros2_control framing URML engaged earlier; your angler (vehicle-manipulator systems) is where URML's manipulation primitives extend underwater. The point of the layer is validate-before-actuate: a command outside the declared depth rating, thruster envelope, or comms regime is refused before it reaches a thruster.

Three real questions: (1) What should a URML capability manifest declare to honestly describe an underwater vehicle — depth rating, thruster/actuator configuration, buoyancy/ballast, tether vs untethered comms, current/visibility limits? (2) Is a validated natural-language intent layer above blue interesting for the lab's BlueROV2 / UVMS work? (3) Where is the cleanest seam — the ROS 2 interface, or the auv_controllers ros2_control layer?

Full write-up: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0396-blue-rdml-outreach.md

Thanks for blue / auv_controllers / angler; a clean, open ROS 2 underwater stack is exactly what a neutral intent layer wants to sit above honestly.

Ido Yahalomi (URML, greenvh@gmail.com)

AI-assisted prose, maintainer-reviewed before posting (see VIBE.md). Human-only correspondence available on request.