Skip to content

URML: a capability check at the sim-to-real deploy boundary #316

Description

@idoco2003

Hi. Up front: this is AI-assisted prose I reviewed before posting (see https://github.com/URML-MARS/URML/blob/main/VIBE.md); say the word if you would rather correspond human-only.

I maintain URML (https://urml.dev), a small Apache-2.0 language that checks an intended action against a robot's declared capability manifest and safety envelope before it runs.

The deploy step of a sim-to-real stack is exactly where that check earns its place: a policy trained in simulation can command an action that is fine in sim and out of bounds on the real arm. RobotControlStack takes a trained policy and runs it on a real Franka, UR5e, or xArm; URML can declare that specific arm's reach, payload, joint limits, and the active keep-out and speed envelope, and validate a commanded action against it at the moment before it leaves for the real robot. It is a static admissibility check at the sim-to-real boundary, not a controller and not part of training.

To be clear about the license: your project is AGPL-3.0, so this is a cross-reference, not a proposal to reuse code either direction.

Two questions:

  1. At the sim-to-real deploy boundary, is a declared-capability and envelope check on the policy's commanded action a useful guardrail before it reaches the real arm, or is that already covered by the deploy path?
  2. Would a small worked example mapping a deployed action onto a URML manifest, validated with no execution, be worth having?

Nothing here asks you to adopt, host, or maintain anything. Thanks for a lean, ROS-free deployment stack.

Ido Yahalomi (greenvh@gmail.com)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions