[FEATURE] Add support of link filtering to contact sensors. - #2772
Merged
duburcqa merged 3 commits intoJul 21, 2026
Merged
Conversation
Member
|
This feature has already been implemented by #2655 . |
Contributor
Author
|
@Milotrince What I'm after here is a |
Milotrince
force-pushed
the
feature/contact-force-filter-link-idx
branch
from
July 8, 2026 00:26
d42940d to
6f1b2ea
Compare
Milotrince
changed the title
Member
|
I rebased your branch and updated to extend the filtering to ContactDepthProbe and KinematicTaxel as well as ContactProbe! |
Milotrince
force-pushed
the
feature/contact-force-filter-link-idx
branch
from
July 8, 2026 01:29
6f1b2ea to
1358dea
Compare
Milotrince
added
the
ready for review
Milotrince
previously approved these changes
Jul 8, 2026
Milotrince
force-pushed
the
feature/contact-force-filter-link-idx
branch
2 times, most recently
from
July 11, 2026 00:06
82e30bd to
37e6118
Compare
|
|
Milotrince
force-pushed
the
feature/contact-force-filter-link-idx
branch
2 times, most recently
from
July 14, 2026 21:19
645ea78 to
d8a6312
Compare
duburcqa
reviewed
Jul 17, 2026
duburcqa
reviewed
Jul 17, 2026
duburcqa
reviewed
Jul 17, 2026
Milotrince
removed
the
ready for review
duburcqa
force-pushed
the
feature/contact-force-filter-link-idx
branch
2 times, most recently
from
July 21, 2026 12:27
c634870 to
ce4e11d
Compare
Collaborator
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ce4e11ddf8
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
duburcqa
force-pushed
the
feature/contact-force-filter-link-idx
branch
from
July 21, 2026 13:43
ce4e11d to
7cc560a
Compare
duburcqa
previously approved these changes
Jul 21, 2026
Collaborator
|
@codex review |
|
Codex Review: Didn't find any major issues. 👍 Reviewed commit: ℹ️ About Codex in GitHubCodex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback". |
ContactForceSensor now accepts filter_link_idx, mirroring ContactSensor: contacts with the sensor link whose other participant is one of the listed links are excluded from the reported force. The filter is applied on both ground-truth-update paths -- the vectorized zerocopy path (same 4D contact-vs-filter broadcast as ContactSensor, applied to the per-side contact masks before they are aggregated) and the _kernel_get_contacts_forces kernel (a small per-contact loop over the sensor's filter list). As in ContactSensor, filtered_sensor_idx gates the more expensive comparison so sensors without a filter keep the cheap path. Covers this in test_contact_sensor_filter_link_idx alongside the Contact sensor.
ContactProbe, ContactDepthProbe, and KinematicTaxel now accept filter_link_idx, mirroring the Contact/ContactForce semantics: a contact with the sensor link is dropped when its counterpart link is in the filter list. The three share the counterpart check via a _func_link_is_filtered device helper threaded into both contact-depth query funcs and their kernels, and a FilterableTactileProbeSensorOptionsMixin so the point-cloud tactile sensors (Elastomer/Proximity), which sample a point cloud rather than solver contacts, do not expose the option. Consolidate the existing contact-force filtering while here: - the kernel now uses the shared _func_link_is_filtered helper instead of open-coding the per-contact filter flags; - the zerocopy filter masking (identical in Contact and ContactForce) is factored into _drop_filtered_counterpart_contacts; - the build-time padding is factored into misc.append_filter_links_idx (used by all five sensors) and the range validation into options._validate_filter_link_idx. The filter table keeps at least one column so the no-filter case is still a valid kernel argument. Covers this in test_kinematic_contact_probe_box_sphere_support.
duburcqa
force-pushed
the
feature/contact-force-filter-link-idx
branch
from
July 21, 2026 14:04
7cc560a to
2107e29
Compare
…aycast path. Builds on the two preceding contributor commits and addresses the review feedback: - Replace the free _validate_filter_link_idx and the separate FilterableTactileProbeSensorOptionsMixin with a single ContactFilterOptionsMixin shared by Contact, ContactForce, and the solver-contact tactile probes, so the generic sensor options no longer reference links. - Replace misc.append_filter_links_idx with a shared ContactFilterMetadataMixin that owns the per-sensor filter table and its fast-path index. - Apply the tactile counterpart filter where each sensor's per-(env, sensor) contact/geom list is built (_kernel_build_sensor_geom_idx for SDF, _kernel_build_sensor_contact_idx for raycast) -- before the per-sensor contact cap, so both contact-depth backends honor it and a large filtered manifold cannot starve an allowed contact, leaving the per-probe hot kernels untouched. - Re-home the tests into tests/sensors/ and add regression coverage, including the prefilter-cap starvation case.
duburcqa
force-pushed
the
feature/contact-force-filter-link-idx
branch
from
July 21, 2026 14:32
2107e29 to
ec68096
Compare
duburcqa
approved these changes
Jul 21, 2026
duburcqa
changed the title
jsw7460
added a commit
to jsw7460/Genesis
that referenced
this pull request
Jul 23, 2026
…sis-Embodied-AI#2772) Genesis upstream added first-class link filtering to the contact sensors (ContactFilterOptionsMixin.filter_link_idx on Contact, ContactForce, and the tactile probes), superseding the local ContactForceSensor filter patch this fork carried. Drop the local patch and take the upstream implementation wholesale: contact_force.py and options.py now match upstream exactly, so the fork no longer diverges and future pulls stop conflicting. The filter_link_idx option name and semantics are unchanged, so downstream callers keep working.
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
3 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.
ContactForceSensor now accepts filter_link_idx, mirroring ContactSensor: contacts with the sensor link whose other participant is one of the listed links are excluded from the reported force. The filter is applied on both ground-truth-update paths; the vectorized zerocopy path and the _kernel_get_contacts_forces kernel.
Adds test_contact_force_sensor_filter_link_idx.
Description
ContactForceSensornow acceptsfilter_link_idx(default()), mirroringContactSensorfrom #2655: contacts with the sensor link whose otherparticipant is one of the listed links are excluded from the reported force.
Related Issue
Resolves #2771
Motivation and Context
ContactSensoralready filters its boolean by counterpart link, butContactForceSensoronly reported the total contact force on the link, with no way to scope it (e.g. "force on a foot from the ground, excluding self-contact").How Has This Been / Can This Be Tested?
test_contact_force_sensor_filter_link_idxputs a box on the floor (env 0) / a box on the floor with another box stacked on top (env 1), with twoContactForcesensors on the bottom box — one plain, one withfilter_link_idx (floor.link_start,). It asserts the filtered sensor reads ~0 in env 0 and reads only the top box's downward push in env 1, while the unfiltered sensor reads the net contact force in both.Checklist:
Submitting Code Changessection of CONTRIBUTING document.