URML (open robot intent language): declarative intent as a teaching on-ramp alongside imperative bots (request for comment)

[idoco2003]

Hi Tank Royale maintainers,

URML (urml.dev) is a small, Apache-2.0 language for robot intent: a goal becomes a typed primitive, validated against the actor's declared capabilities, then dispatched. Tank Royale is a programming game widely used to teach programming, and while a game bot is virtual, the shared idea is interesting: a declarative, checkable way to state what a bot should try to do. This is a conceptual note, not an integration ask.

Nothing here asks the project to adopt, host, or maintain anything. This is a request for comment.

The idea: a Tank Royale bot is written imperatively against an API. URML's angle is a declarative, typed intent layer that states a goal and is validated against the actor's declared capabilities. For a teaching game, an optional declarative-intent mode could be a gentle on-ramp before learners write full imperative bots, and it makes "what is this bot allowed to do" explicit. This is exploratory; a battle game is not a physical robot, and the value, if any, is pedagogical.

Two real questions: (1) is a declarative, typed intent layer an interesting on-ramp or teaching aid alongside imperative bot programming? (2) Does "declare what the bot should try to do, validated against its capabilities" map onto how Tank Royale bots are written?

Full write-up: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0574-tank-royale-outreach.md

Thanks for Tank Royale; carrying the Robocode tradition forward is great, and the declarative-on-ramp idea felt worth floating with you.

Ido Yahalomi (URML, greenvh@gmail.com)

AI-assisted prose, maintainer-reviewed before posting (see https://github.com/URML-MARS/URML/blob/main/VIBE.md). Human-only correspondence available on request.

[flemming-n-larsen]

@idoco2003

Hi Ido, thanks for the thoughtful note and the kind words about Tank Royale.

One thing worth pointing out: Tank Royale already has a declarative intent layer at its core. Every turn a bot sends a BotIntent (target speed, turn rates, gun turn rate, firepower), and the server validates and clamps each value against the bot's capabilities (max speed, turn-rate limits, gun heat) before applying it. So "declare what the bot wants to do, validated against its capabilities" is essentially the existing protocol, just at the per-turn physics level rather than at a goal level.

A higher-level goal layer ("attack", "evade", etc.) is a different thing, closer to behavior trees or goal-oriented action planning. It's a reasonable pedagogical idea, but it doesn't map naturally onto how bots are actually written here, which is event-driven and reactive (onScannedBot, onHitByBullet, and a tight per-turn loop). A declarative goal mode would sit on top of that loop rather than replace it, and anyone is free to prototype such a layer on top of the existing Bot API.

For the project itself this is out of scope. Tank Royale is maintained solo, and the imperative, event-driven API is intentionally the foundation. Appreciate you floating the idea, and good luck with URML.

Flemming

[idoco2003]

@flemming-n-larsen Thanks Flemming, the BotIntent distinction is a useful one. Good to know the per-turn validated-intent layer is already there; I hadn't dug that deep into the protocol. Understood that a higher-level goal layer is out of scope for the project.
Good luck with Tank Royale.