Hi Forge maintainer,
URML (urml.dev) is a small, Apache-2.0 language for robot intent: an intent becomes a typed primitive, validated against the robot's declared capabilities and a safety envelope, then dispatched. Forge converts between robot-learning data formats (RLDS, LeRobot, rosbag), and URML is interesting as an annotation source for those datasets.
Nothing here asks the project to adopt, host, or maintain anything. This is a request for comment.
The mapping: a robot-learning episode records what happened. URML's audit trail records the typed intent that drove it and the validation verdict. Aligning the two gives episodes a structured intent label without hand-annotation -- exactly the kind of label robot-learning datasets are usually missing. Forge already moves between formats; URML intent records could ride along as an annotation channel. The intent vocabulary is small and typed, so the annotation stays compact and consistent across robots and substrates.
Two real questions: (1) is a typed validated-intent record a useful annotation channel when converting robot-learning datasets? (2) Does aligning intent records to RLDS/LeRobot/rosbag episodes fit Forge's model?
Full write-up: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0555-forge-outreach.md
Thanks for Forge; format conversion is the natural place to thread a consistent intent annotation through the robot-learning data world.
Ido Yahalomi (URML, [email protected])
AI-assisted prose, maintainer-reviewed before posting (see https://github.com/URML-MARS/URML/blob/main/VIBE.md). Human-only correspondence available on request.
Hi Forge maintainer,
URML (urml.dev) is a small, Apache-2.0 language for robot intent: an intent becomes a typed primitive, validated against the robot's declared capabilities and a safety envelope, then dispatched. Forge converts between robot-learning data formats (RLDS, LeRobot, rosbag), and URML is interesting as an annotation source for those datasets.
Nothing here asks the project to adopt, host, or maintain anything. This is a request for comment.
The mapping: a robot-learning episode records what happened. URML's audit trail records the typed intent that drove it and the validation verdict. Aligning the two gives episodes a structured intent label without hand-annotation -- exactly the kind of label robot-learning datasets are usually missing. Forge already moves between formats; URML intent records could ride along as an annotation channel. The intent vocabulary is small and typed, so the annotation stays compact and consistent across robots and substrates.
Two real questions: (1) is a typed validated-intent record a useful annotation channel when converting robot-learning datasets? (2) Does aligning intent records to RLDS/LeRobot/rosbag episodes fit Forge's model?
Full write-up: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0555-forge-outreach.md
Thanks for Forge; format conversion is the natural place to thread a consistent intent annotation through the robot-learning data world.
Ido Yahalomi (URML, [email protected])
AI-assisted prose, maintainer-reviewed before posting (see https://github.com/URML-MARS/URML/blob/main/VIBE.md). Human-only correspondence available on request.