A small, opinionated, human-readable language for describing robot intent.

urml.dev

---

RFC-0470: BehaviorTree.CPP integration — request for comment

Kind: Outreach. No spec change is proposed here. This is a per-target request for comment proposing how URML v0.1 maps onto an existing target and asking its maintainers for feedback. It builds on URML's Layer-3 behavior composition. It is the anchor of the behavior-tree / orchestration wave (Move #41).

Summary

BehaviorTree/BehaviorTree.CPP (MIT, ~4.1k stars, active) is the dominant C++ behavior-tree library in robotics (it is what Nav2 runs on), with BehaviorTree.ROS2 wrapping action/service/topic leaf nodes. URML has a Layer-3 of its own — programs are trees of sequence / parallel / branch / retry over typed, validated primitives. The two meet cleanly: URML is the validated-intent layer, a behavior tree is the execution engine. This RFC asks how they should interop.

The mapping (URML above / beside BehaviorTree.CPP)

Two complementary seams, both honest:

• URML lowers to a tree. A validated URML program (its sequence/parallel/branch/retry structure) compiles to a BehaviorTree.CPP tree; each URML primitive becomes a leaf. URML's contribution is what BT leaves do not check on their own: the typed args, the capability match, and the safety envelope, all verified before the tree runs.
• A leaf dispatches a validated primitive. Alternatively a custom BT.CPP/BT.ROS2 node wraps a single URML primitive (move_to, grasp, set_output), so a hand-authored tree gets validate-before-actuate per leaf.

The acid test holds: a behavior tree is control flow; URML is the typed, capability-checked intent the flow carries.

What is asked

Request for comment from the BehaviorTree.CPP maintainers:

1. Which seam is more natural — URML compiling to a BT.CPP tree, or a BT leaf node that dispatches a validated URML primitive?
2. Does URML's sequence/parallel/branch/retry map cleanly onto BT.CPP control nodes (Sequence/Parallel/Fallback/RetryUntilSuccessful), or are there mismatches?
3. Is a validated-intent leaf node interesting for the BT.ROS2 node set?

Nothing here asks the project to adopt, host, or maintain anything.

Prior art / context

URML's Layer-3 behavior composition (sequence/parallel/branch/retry, RFC-0002); the manipulation/mobility primitive families. BehaviorTree.CPP is the anchor of the behavior-tree / orchestration wave; the sibling BehaviorTree.ROS2 is referenced, not posted to separately (org-anchor).

Implementation note

Outreach only. The post is a GitHub Discussion on BehaviorTree/BehaviorTree.CPP under the maintainer's identity, AI-assisted-authoring disclosure (VIBE.md) up front, no license-ask (MIT). BehaviorTree.ROS2 is referenced, not posted to separately. Tracked in examples/lighthouses/outreach-move41.yaml.