Repository navigation
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds a slow test lift, continued grip monitoring, measured-joint holding on alarm and one supported open/recenter/retry. Original mechanics, force limits, free-object dynamics and the task scorer remain unchanged.
The complete fresh 30-attempt confirmation is committed under
benchmarks/piper-grip-20261011/receipt. All attempts are numerically valid, but the guard regresses placement completion: original 4/6 versus guarded 2/6. Both pass normal friction; at intermediate friction original passes 2/2 while the guard aborts both. Low-friction attempts retry once each and abort without qualification. No successful recovery is demonstrated. All 18 negative controls fail to complete, as expected.The likely over-conservative displacement threshold is discussed alongside the need for an identical-motion/no-abort comparison. Short probe qualification is separate from eventual completion; sampled loss proxies are not verified physical drops. All attempts, failures, commands, states, stored/decoded hashes and timing outliers remain available.
Validation: 44 focused native/controller/benchmark tests and 30 publisher/preparer tests pass. An installed wheel matched every frozen runtime source hash and executed with network denied. The full 600 simulated seconds plus compression and final inspection took 104.261 wall seconds on the recorded CPU host; 53 of 30,000 control windows exceeded 20 ms. This is not a sensor-to-actuator latency claim.
The authored controller uses privileged simulator observations; no learned policy or physical robot qualification is claimed. Frozen runtime:
3ee94706f436690c6d1d8554eda9202d23620e3c, seed 731209.This PR is stacked on #11 (
feat/reference-task-benchmark). The documented candidate branch is installable. Research article · Hugging Face lab.