feat(simulation): opt-in ground-truth object pose stream for eval scoring - #3960
Open
Jerrybery wants to merge 1 commit into
Open
feat(simulation): opt-in ground-truth object pose stream for eval scoring#3960Jerrybery wants to merge 1 commit into
Jerrybery wants to merge 1 commit into
Conversation
…ring MujocoSimModule gains publish_ground_truth (default off) and ground_truth_hz (20 Hz). When enabled, world poses of every free-joint scene body (robot root excluded — odom already covers it) publish on a new gt_object_poses Out[PoseStamped] stream, frame_id = body name. Ground truth is privileged scoring data for the evals framework (issue dimensionalOS#3594): never consumed by the agent. Part 1 of 2; part 2 adds the eval-side gt_store() and predicate library.
Jerrybery
requested review from
Dreamsorcerer,
aclauer,
leshy,
mustafab0,
paul-nechifor and
spomichter
as code owners
September 7, 2026 05:59
github-actions
Bot
added
the
first-time-contributor
Contributor
Greptile SummaryThis change adds an opt-in, throttled ground-truth pose stream for eligible free-joint scene bodies. It is not safe to merge until Confidence Score: 4/5Not safe to merge because an enabled ground-truth stream can crash from an accepted configuration value. Execution reproduced the failure in the ground-truth publishing path and also demonstrated incorrect behavior for other invalid rates. Files Needing Attention: dimos/simulation/engines/mujoco_sim_module.py
 T-Rex LogsWhat T-Rex did
Important Files Changed
|
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| # free-joint scene bodies (robot root excluded, odom covers it). Never | ||
| # consumed by the agent. Off = zero overhead, zero behavior change. | ||
| publish_ground_truth: bool = False | ||
| ground_truth_hz: float = 20.0 |
Contributor
There was a problem hiding this comment.
When ground-truth publishing is enabled, ground_truth_hz=0 reaches the throttle calculation and raises ZeroDivisionError in the post-step callback. Negative, infinite, and NaN rates are also accepted, causing immediate or permanently suppressed publishing instead of the requested stream. Require a finite value strictly greater than zero during configuration validation.
Artifacts
- This authored harness extracts and runs the unchanged production post-step method with enabled ground-truth publishing and invalid rates, showing the affected execution path.
- This command capture shows zero raises ZeroDivisionError, while negative and infinity publish and NaN suppresses publication, confirming invalid values are not safely handled.
❌ 3 Tests Failed:
View the top 3 failed test(s) by shortest run time
To view more test analytics, go to the Test Analytics Dashboard |
Author
|
Part 2 of the #3594 split is implemented and pushed as a stacked branch on my fork: It adds the eval-side consumption of this PR's GT stream:
I'll open it as a PR once this one lands. Local verification: Note: the current |
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
1 participant
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.
Summary
Part 1 of 2 for #3594 (privileged ground-truth channel for interactive eval scoring).
MujocoSimModulegains an opt-in ground-truth stream:publish_ground_truth: bool = False(default off — zero overhead, zero behavior change) andground_truth_hz: float = 20.0config fieldsgt_object_poses: Out[PoseStamped]stream: world poses of every free-joint scene body, throttled,frame_id= body nameodom); resolved by matching the same root qpos addressodomuses/from MuJoCo 3.10's attached-submodel namespacing is stripped soframe_idstays a stable plain body nameGround truth is privileged scoring data — never consumed by the agent. Part 2 will add the eval-side
gt_store()accessor, a predicate library, and an xArm7 regression suite.Test plan
test_mujoco_sim_module.py: disabled-by-default publishes nothing; enabled publishes throttled poses; free-joint body resolution excludes the robot root against a real composed modelNote for part 2 consumers:
async def handle_<stream>auto-subscription uses a single-slot LATEST mailbox (module.py), so same-tick bursts overwrite; GT consumers must use directsubscribe()(asRecorderdoes).