You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
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. Autoware is named in URML's manifesto as a target, and URML's research-grade autonomous-vehicle profile was designed against Autoware's operational model, which is why I am writing.
Nothing here asks Autoware to adopt, host, or maintain anything. This is a request for comment, and genuinely a design question.
URML sits one layer above Autoware as a substrate-neutral, validated intent vocabulary. The mapping I shipped (urml.dev, RFC-0020):
plan_path(from, to, along) is a compute verb that binds a planned trajectory; it maps onto the planning pipeline (mission -> behavior -> motion). It does not actuate.
follow_trajectory(route, speed_envelope) is the only verb that actuates; it maps onto the control stack (pure_pursuit / MPC). The validator checks the trajectory's speed against the declared ODD cap, and that follow_trajectory consumes a plan_path-produced trajectory, before anything dispatches.
A manifest av block declares the HD map (format-neutral; Lanelet2 is yours), the ODD, and the Minimum-Risk Maneuver.
URML is explicitly research-grade for AV (production_safety_certified: false). It does not certify autonomous-vehicle safety and makes no SOTIF / R157 claim; the value is an honest, validated, human-readable intent layer with a validate-before-actuate guarantee.
Three real questions: (1) Is the plan_path -> planning, follow_trajectory -> control mapping faithful to Autoware's architecture, or does it mis-model the mission/behavior/motion split? (2) What should a URML av manifest declare to honestly describe an Autoware deployment's ODD beyond a speed cap and regions? (3) Where is the right seam for a URML -> Autoware adapter: the trajectory ROS 2 interface, or higher?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
Hi Autoware community,
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. Autoware is named in URML's manifesto as a target, and URML's research-grade autonomous-vehicle profile was designed against Autoware's operational model, which is why I am writing.
Nothing here asks Autoware to adopt, host, or maintain anything. This is a request for comment, and genuinely a design question.
URML sits one layer above Autoware as a substrate-neutral, validated intent vocabulary. The mapping I shipped (urml.dev, RFC-0020):
URML is explicitly research-grade for AV (production_safety_certified: false). It does not certify autonomous-vehicle safety and makes no SOTIF / R157 claim; the value is an honest, validated, human-readable intent layer with a validate-before-actuate guarantee.
Three real questions: (1) Is the plan_path -> planning, follow_trajectory -> control mapping faithful to Autoware's architecture, or does it mis-model the mission/behavior/motion split? (2) What should a URML av manifest declare to honestly describe an Autoware deployment's ODD beyond a speed cap and regions? (3) Where is the right seam for a URML -> Autoware adapter: the trajectory ROS 2 interface, or higher?
Full write-up: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0386-autoware-outreach.md
Thanks for Autoware; an open, production-aimed AV stack is exactly the kind of substrate a neutral intent layer should be able 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.
Beta Was this translation helpful? Give feedback.
All reactions