Skip to content

URML (open robot intent language): bounding a legged_gym-trained locomotion policy, request for comment #98

@idoco2003

Description

@idoco2003

Hi legged_gym maintainers,

URML (urml.dev) is a small, Apache-2.0 language for describing robot intent: it validates a request against a capability manifest and a safety envelope, then dispatches. legged_gym is where a remarkable number of quadruped and humanoid walk policies are trained, and a trained policy is exactly the kind of thing URML would treat as a locomotion substrate: URML declares the platform's capability and bounds a navigation-level intent the policy must satisfy, while the policy owns the gait.

Nothing here asks legged_gym to change or maintain anything. This is a request for comment.

Two real questions. First, what is the right URML granularity above a velocity-tracking locomotion policy: navigation-level commands (go to X, walk at velocity V, turn), or something finer? Second, URML has no legged mobility class today; for a platform whose locomotion is a legged_gym-trained policy, what should the manifest declare as capability versus leave to the policy? I am reaching the rsl_rl maintainers separately about the library underneath.

Full write-up: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0376-legged-gym-outreach.md

Thanks for legged_gym; it is the on-ramp for a huge amount of legged-locomotion work.

Ido Yahalomi (URML, greenvh@gmail.com)

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    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