Repository navigation
feat(robotd): execute timed action chunks for VLA runners - #1
Merged
Merged
Conversation
A model runner can now submit a timed trajectory directly to robotd instead of translating every prediction into predefined movement skills. The existing 50 Hz motor loop admits and executes bounded 15-joint absolute-radian chunks through Safety and RobotIo; model inference and action decoding remain outside the daemon. Add begin/submit/end JSON-RPC calls, robotctl commands and robot.state execution feedback. Robot-clock scheduling skips expired samples, replaces future trajectory suffixes, rejects stale sessions and competing joint writers, and invalidates queued results on operator cancellation. Exhaustion or body failure releases ownership into measured-pose hold without automatically resuming the locomotion policy. Shared API version is now 38. This entrypoint is enabled only on --fake/--sim at 50 Hz and is accessed through the local Unix socket. Physical hardware and BLE/WebRTC forwarding are excluded from this initial integration. No VLA weights, inference service or learned balance behavior are included. Validation: - Rust 1.99 Linux: cargo test --workspace --locked --offline — 1550 passed, 18 ignored, 0 failed. - cargo fmt --all -- --check and git diff --check passed. - Real robotd/FakeIo IPC regression covers measured motion, conflicting commands, stop and rejection of late results. - Isolated MuJoCo body: head_yaw moved from about 0.024 to 0.142 rad toward a 0.174 rad target; cancellation entered held state and rejected the old session. This checks joint execution, not balance; the unbalanced body fell with the locomotion policy disabled. Assisted-by: Codex
Coverage73.17% lines on this branch, against a floor of 70%. Per-file |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A model runner can now submit a timed trajectory directly to robotd instead of translating every prediction into predefined movement skills. The existing 50 Hz motor loop admits and executes bounded 15-joint absolute-radian chunks through Safety and RobotIo; model inference and action decoding remain outside the daemon.
Add begin/submit/end JSON-RPC calls, robotctl commands and robot.state execution feedback. Robot-clock scheduling skips expired samples, replaces future trajectory suffixes, rejects stale sessions and competing joint writers, and invalidates queued results on operator cancellation. Exhaustion or body failure releases ownership into measured-pose hold without automatically resuming the locomotion policy. Shared API version is now 38.
This entrypoint is enabled only on --fake/--sim at 50 Hz and is accessed through the local Unix socket. Physical hardware and BLE/WebRTC forwarding are excluded from this initial integration. No VLA weights, inference service or learned balance behavior are included.
Validation:
Assisted-by: Codex