Repository navigation
decisions: shared base, embodied-AI repo, platform-sw pin, ToF pitch 10 deg, base_link height in FRAMES-REP105 - #55
Conversation
…10 deg, base_link height in FRAMES-REP105 - SHARED-BASE (new): one base electrical design for release 001 and 2.0; only the base controller and sensor types differ; WCM-500-24 on both. - CHARGING: release 001 uses the same WCM-500-24 path. - EMBODIED-AI-REPO (new): recorder, policy runner, dataset QA, ACT baseline and evaluation in openamrobot-embodied-ai; interfaces keeps the C9 schema; datasets and checkpoints on the Hugging Face Hub. - PLATFORM-PIN (new): the combined model is tested against one pinned platform-sw SHA; the pin moves only with a full A6/A7 rerun. - NEAR-FIELD-FRAMES: ToF cliff pitch 10 deg down (was 20 deg screening); positions unchanged until the robot-description owner confirms CAD values. - FRAMES-REP105: base_link 0.098 m above the floor moved here from CAMERAS. - Tests for every new and changed pattern; WATCHDOG.md accepted-words table regenerated. Signed-off-by: Alex Reznichenko <info@botshare.ai> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015uY6hFMbCNTirChFEieXEs
panthera-momagdii
left a comment
panthera-momagdii
left a comment
There was a problem hiding this comment.
Reviewed 8a180c395ca6b14e0fbf32adf0bdb935a23db011 against 9109d3e1a8fc07151e500e83b32b7ce6ccab6af6, including all changed files, #54, and the current relevant consumers.
The repository split, fixed platform-pin policy and relocation of the unchanged base-link height are consistent with the inspected sources. The intended shared design remains distinct from the installed legacy charging unit. Software review does not replace the scoped electrical/model owner approvals named in #54.
Two corrections are needed before approval: link the complete ToF consumer migration, including the executable model, and make the outstanding bench qualification unambiguous. Details are inline. Recording a chosen 10-degree design does not itself establish physical qualification.
Verification: hosted run 39430232439 reports 271 tests passed and register validation passed; 80 focused decision-checker tests also passed independently in an exact-source snapshot. Selected consumer scans demonstrate the outstanding mismatches. These checks establish textual consistency, not robot geometry or bench acceptance.
ChatGPT-assisted source, CI and isolated checker review; no ROS, simulation or hardware execution.
| date: 2026-10-20 | ||
| review_by: 2026-12-01 | ||
| source: {document: Alex decision 2026-10-20, item: "ToF cliff pitch 10 degrees; frame positions from P-03 rev19.0 item 19, unchanged until the robot-description owner confirms the CAD values"} | ||
| supersedes: |
There was a problem hiding this comment.
[P2] Link the complete ToF consumer migration
#54 lists the architecture prose and sensor_frames.yaml for the pitch change, but the executable sensors.xacro and robot.sdf still use 0.3491 rad (20 degrees). The hardware near-field document also retains 20 degrees and is missing from the migration list. Updating only the listed prose/datum files would leave the actual model inconsistent. Please link owner-reviewed consumer PRs covering these files, the datum table and the software documentation, as required by the decision-change rule. The new text patterns do not detect these numeric model values, and SDF is not in this entry's scanned file types; a green checker alone will not close this gap.
| summary: "Four MB7060 frames us_* at x +/-0.390, y +/-0.130; two TF-Luna frames tof_* at x +/-0.390 on the centreline." | ||
| fix_hint: "Use the listed frame origins; the ToF pitch of 20 degrees is a screening value until the bench test." | ||
| summary: "Four MB7060 frames us_* at x +/-0.390, y +/-0.130; two TF-Luna tof_* at x +/-0.390, pitched 10 deg down." | ||
| fix_hint: "Use the listed frame origins; the ToF pitch is 10 degrees down, confirmed in the bench test." |
There was a problem hiding this comment.
[P2] Keep bench confirmation explicitly pending
This fix_hint is printed by the checker and currently tells users the 10-degree pitch is 'confirmed in the bench test'. No completed bench result is linked, while this entry's verification text still treats premature validation claims as an error. Please say that 10 degrees is the chosen design pending a real-floor return test, or link the actual completed qualification if available. Apply the same explicit status to the values/PR summary. The floor-intersection geometry is a calculation, not evidence of a successful sensor return on real floors.
…ion owner - Ultrasonic frames us_* stay at x +/-0.390, y +/-0.130, z -0.0066, now confirmed flush with the panel outer face; the CAD offsets +0.6 / -1.4 mm only reflect the sensor depth in the lock nut. - ToF frames: tof_front (0.399, 0, -0.0066) and tof_rear (-0.399, 0, -0.0066, yaw pi), lens face 9 mm ahead of the panel on mount MMP.00.00.00.005, pitched 10 degrees down. "tof_* at x +/-0.390" moves to supersedes (P-03-rev19.0 item 19 D6). - Source: Alex decision 2026-10-20 (ToF pitch) and robot-description owner confirmation 2026-10-20 (frame x values); y and height from P-03 rev19.0 item 19. - The summary drops "frames" and "pitched" to stay within the 120-character limit. No check pattern or generated table changes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TPiK6pUmmjNhPceECkKpR2 Signed-off-by: Alex Reznichenko <info@botshare.ai>
|
NEAR-FIELD-FRAMES updated to the owner-confirmed values: ultrasonic x ±0.390 at the panel face, ToF x ±0.399 at the lens face, ToF pitch 10 deg. |
Records the platform decisions of 19 and 20 Oct 2026 in the register. Change request: #54.
Changes
Evidence
git amto main 9109d3e without changes (commit 8a180c3, original author and sign-off kept).python3 tools/check_decisions.py --decisions decisions.yaml --maintainers maintainers.yaml --validate-only: passed (39 loaded, 35 recorded and scanned, 1 open, 3 superseded).python3 -m unittest discover -s tests: 271 tests, all passed (OK). The WATCHDOG.md accepted-words table needed no regeneration.Follow-up PR: M07-REST-STATE (after @Brinex45 confirms on openamr-upperbody-sw #14) and the NEAR-FIELD-FRAMES positions (after the robot-description owner confirms).