[Vertax42]

Summary

Two commits:

1. feat(firmware): ship EL05 motor firmware in firmware/, pick it by recorded model

   • firmware/EL05_1.0.5.0.4.bin added and listed under a new motors section of firmware/manifest.json (CRC32 0x845a0474).
   • motor_ota_update.py image argument is now optional: without one it uses the shipped image for the motor model recorded on the gripper. It refuses when only the compile-time default model is known, because RobStride's OTA accepts any model's image.
   • The shipped image is checked against the manifest CRC32; a bare image name also resolves under firmware/. Naming an image explicitly works as before.
   • No RS00 image ships yet.
   • New test python/tests/test_firmware_manifest_motors.py; docs (firmware/README.md, docs/MOTOR_OTA*.md, docs/EXAMPLES.md) and CHANGELOG updated.

2. refactor: translate all Chinese in source code to English

   • Comments, docstrings, pybind11 docstrings and runtime text (logs, exceptions, example output, argparse help) across cpp/, python/, examples and tests. No logic changes.
   • Runtime text that tests match on changed with them (test_power_cycle_warnings_pty: "below 1.2.11"). Bilingual messages in follower_gripper.cpp keep one full English version. Example table headers use short English labels at the same widths.
   • Anything that filters logs on the old Chinese text needs updating (noted in CHANGELOG).

Test plan

• C++: 471/471 ctest pass
• Python: 148 passed, 2 skipped (taccap env, pybind11 from the env)
• ruff check / ruff format --check clean
• No CJK characters left in tracked non-Markdown files
• Motor OTA on a real EL05 follower without an image argument

🤖 Generated with Claude Code

[xensexyq]

Retrospective review and validation report

Adding the local test and physical-device acceptance results from 2026-10-20 (Asia/Shanghai) for the author. These checks were completed before merge; this comment records that evidence after merge and does not represent another flashing run.

• Reviewed/tested head: 3b604da; original base: fd0d3a5.
• Merge: 29ac807. Its tree matches the tested head. The isolated test source was also rechecked against all 170 tracked files in that head.

Software validation

Environment: Linux x86_64, CPython 3.12.14, GCC 14.4.0 and OpenCV 4.12.0. The native SDK was built from the isolated PR snapshot; its actual import path was checked to avoid accidentally testing the editable main checkout.

Check	Observed result
Release build of the C++ core and Python bindings	PASS
Full Python suite	148 passed, 2 skipped; both skips are leader-only IMU cases because no leader was connected
OTA-related tests (test_firmware_manifest_motors.py, test_motor_ota_update.py, test_ota_update_all.py)	27 passed, already included in the full-suite count
PR CI	build + unit tests, lint and protocol mirror vs firmware: SUCCESS

The CI results above are the PR's recorded pre-merge checks, separate from the local test run.

Physical EL05 acceptance

Target: one right-hand EL05 follower (SN ending 0016s), MCU firmware 1.2.16. The selected image was the PR's bundled firmware/EL05_1.0.5.0.4.bin, chosen from the model recorded on the follower without supplying an image path.

Image identity: 92,584 bytes, CRC32 0x845a0474, SHA-256 fb232707a852f9144bdf4a744ee9c89d39c8d96695189f61ff1e6b8eac293f0f.

Procedure: run python python/examples/motor_ota_update.py <follower-SN> for stage 1, cycle 24 V for about 2 seconds and wait about 10 seconds, run the same command for stage 2, then cycle 24 V again and wait for homing. The test used a guarded launcher to ensure these commands loaded the PR-built native module. The operator confirmed both power cycles and final homing, with the jaws clear.

Step / acceptance check	Observed result
Stage 1: persist MIT → private protocol selection	PASS
Stage 2: transfer the bundled motor image	PASS; 46.5 s. The log contained one initial OTA-start timeout, after which the transfer completed successfully
Live motor firmware read before/after transfer	1.0.5.0.4 → 1.0.5.0.4; version recorded in follower flash
Motion envelope	Already at the motor rating; unchanged
After second power cycle	MIT confirmed; EL05 model and motor-version flash records valid
Motor / latched / firmware fault codes	0 / 0 / 0
follower_status.py <follower-SN> --seconds 10 --hz 100	1,000 frames / 10.01 s = 99.9 Hz
Status freshness	Firmware timestamp advanced 10,007 ms, matching 10,007 ms wall time; PASS

After merging, main was synchronized and the SDK was rebuilt/reinstalled in the existing environment. A separate import check outside the repository, without PYTHONPATH or LD_LIBRARY_PATH, returned taccap-gripper OK; version 0.4.3; the Python source and installed native-library paths were verified.

Review conclusion and limits

No confirmed blocking defects were identified in the reviewed scope. The recorded checks support bundled-image selection and the two-stage motor OTA/restart path on this EL05 follower. There are no confirmed code findings requiring inline comments in this report.

This was a same-version reflash, not an upgrade from an older motor firmware. This run did not exercise MCU OTA, RS00, Jetson hardware, interrupted-transfer recovery, or a post-upgrade commanded-motion/load test. Leader IMU acceptance remains untested in this run. Local C++ unit tests were not rerun on 2026-10-20; the separate CI result is linked above.

These results should not be generalized to those untested paths. Essential observations are included here so the report does not depend on local-only log paths.