feat(firmware): 随包从爪固件 1.2.15(RS00 按 SN 左右定开合方向)

CHANGELOG.md

@@ -21,6 +21,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   `expected_open_reverse`) is the rule; EL05 is not covered (one unit
   measured).
 
+### Changed
+
+- Shipped follower image is now **1.2.15**: an RS00's open direction follows
+  the side in its firmware SN, auto-calibration corrects a stored direction that
+  disagrees, and a model-record change keeps the side's direction. Verified on
+  0089s (left) / 0094s (right), including a deliberately reversed 0089s
+  corrected on the next 24 V power cycle. Leader image unchanged (1.2.6).
+
 ## [0.4.2] - 2026-10-08
 
 ### Fixed

README.md

@@ -126,7 +126,7 @@ worked. The bank-swap reboot is a soft reset that leaves the device looking heal
 while quietly dropping status frames.
 
 **The two roles carry independent version numbers.** At the time of writing the
-leader is 1.2.6 and the follower 1.2.14; neither is behind the other, and
+leader is 1.2.6 and the follower 1.2.15; neither is behind the other, and
 `gripper.firmware_version` returning different numbers for the two halves of a
 pair is normal. Compare versions only within a role — the floors above are
 follower numbers. **A leader reporting 1.2.6 may be either of two images**: the
@@ -309,8 +309,9 @@ normalized position is mirrored, a closed jaw reads ~1.0 and `set_target(0.0)`
 opens it. `FollowerGripper` warns on open when the stored bit disagrees with
 the side; `t.expected_open_reverse(model_id, side)` gives the right bit. Fix it
 with `g.set_gripper_config()` (change the flags only) and cut 24 V so
-calibration re-runs. A model-record change resets the bit to the model default
-again. EL05 has no rule yet.
+calibration re-runs. Follower 1.2.15 does this by itself: calibration takes the
+direction from the SN's side and corrects a stored bit that disagrees, and a
+model-record change keeps the side's direction. EL05 has no rule yet.
 
 ### By task
 

README_CN.md

@@ -108,7 +108,7 @@ USB 线可以不拔 —— 从爪的 MCU 和电机都靠 24V 运行,只拔 USB 
 主爪没有 24V:拔插它的 USB 即可。`gripper.device.heartbeat().uptime_ms` 从接近 0
 重新计起,就说明真的断过电了。bank-swap 重启只是一次软复位,它会让设备看上去一切正常,却在悄悄丢状态帧。
 
-**主从两个角色的版本号是各自独立的。** 写这段时主爪是 1.2.6、从爪是 1.2.14,
+**主从两个角色的版本号是各自独立的。** 写这段时主爪是 1.2.6、从爪是 1.2.15,
 谁也不比谁旧;一对夹爪的两半 `gripper.firmware_version` 读出不同的数是正常的。
 **只在同一个角色内部比较版本号** —— 上面那些门槛都是从爪的数。**报 1.2.6 的主爪
 可能是两个不同的镜像**:当前的这个带控制串口接收修复;两个角色曾被强行同号那段
@@ -264,8 +264,8 @@ python python/examples/force_position_control.py left --grasp-torque 1.1
 方向原样写回,所以新的左手 RS00 第一次标定一定标反、而且不会自己纠正 —— 开度整个
 颠倒，合着读成接近 1,`set_target(0.0)` 会张开。`FollowerGripper` 打开时若存的方向
 与左右不符会告警;`t.expected_open_reverse(model_id, side)` 给出应有的值。修法：用
-`g.set_gripper_config()` 只改 flags,再拔插 24V 让标定按新方向重跑。改型号记录会把
-它重置回型号默认。EL05 暂无规则。
+`g.set_gripper_config()` 只改 flags,再拔插 24V 让标定按新方向重跑。从爪 1.2.15 起固件
+自己会做：标定按 SN 左右定方向、纠正存反的方向，改型号记录也保留左右方向。EL05 暂无规则。
 
 ### 按任务分
 

docs/FIRMWARE.md

@@ -82,7 +82,7 @@ python python/examples/ota_update.py \
 ```
 
 **The two roles have independent version numbers** — at the time of writing the
-leader is 1.2.6 and the follower 1.2.14, and neither is "behind" the other. Both
+leader is 1.2.6 and the follower 1.2.15, and neither is "behind" the other. Both
 carry the same control-UART receive fix, which lives in shared code.
 The roles were briefly forced onto one number; that was dropped on 2026-09-24.
 

firmware/README.md

@@ -7,7 +7,7 @@ of this SDK.
 | Image | Role | Version | Protocol | Size | CRC32 |
 | --- | --- | --- | --- | --- | --- |
 | `tc-gu-01-master-1.2.6.bin` | leader (SN ends **`m`**) | **1.2.6** | V2.6 | 118,300 B | `0x350393b5` |
-| `tc-gu-01-slave-1.2.14.bin` | follower (SN ends **`s`**) | **1.2.14** | V2.8 + 运动安全包络 | 167,168 B | `0x267d7f67` |
+| `tc-gu-01-slave-1.2.15.bin` | follower (SN ends **`s`**) | **1.2.15** | V2.8 + 运动安全包络 | 167,464 B | `0xbdbb2b4b` |
 
 Only the current release is kept here. Older images come from this directory's
 git history rather than from extra files.
@@ -35,8 +35,8 @@ sources bumps that role alone.
 this directory's `tc-gu-01-master-1.2.6.bin` if you are not sure. See
 docs/FIRMWARE.md.
 
-The leader is at 1.2.6 while the follower is at 1.2.14 for that reason: 1.2.7,
-1.2.8, 1.2.10, 1.2.12, 1.2.13 and 1.2.14 touched only follower code, and 1.2.9 changed the shared
+The leader is at 1.2.6 while the follower is at 1.2.15 for that reason: 1.2.7,
+1.2.8, 1.2.10, 1.2.12, 1.2.13, 1.2.14 and 1.2.15 touched only follower code, and 1.2.9 changed the shared
 `storage.c`, which is what moved the leader from 1.2.4 to 1.2.5 — with no change
 in its behaviour.
 
@@ -76,6 +76,11 @@ the direction are reset to the new model's defaults; a stored calibration torque
 still at exactly the old 0.35 migrates on boot. A gripper OTA also no longer
 refuses when the motor is not on the bus.
 
+**Follower 1.2.15 sets an RS00's open direction from its side.** The gear train
+is mirror-mounted, so a left follower (odd SN) needs Reverse 1 and a right one
+(even) Reverse 0; auto-calibration now uses that and corrects a stored bit that
+disagrees. Requires the SN to be burned correctly. Verified on 0089s / 0094s.
+
 **Follower 1.2.14 records the motor's firmware version in flash** (0x5C to
 write it, 0x58 falls back to it under MIT) -- see the SDK README, *RobStride
 motor firmware*. Verified on 0094s: written, read back under MIT, kept across a

firmware/manifest.json

@@ -13,16 +13,16 @@
       "note": "Leader line. Same fix as follower 1.2.11 in the shared receive path: the control UART (3 Mbaud, one interrupt per byte) stopped receiving for good after a single overrun, because the error callback never restarted reception -- the stream kept flowing while every command went unanswered. Reception is now restarted, and GET_UART_STATS reports rx_errors / rx_rearms. Verified on 0115m and 0116m (upgraded from 1.2.2): 15 min each of a 100 Hz encoder+IMU stream with a 100 Hz acknowledged command, ~87,000 commands, zero failures. AFTER FLASHING, unplug and replug the leader's USB cable."
     },
     "slave": {
-      "file": "tc-gu-01-slave-1.2.14.bin",
-      "version": "1.2.14",
-      "source_commit": "1336ebb",
-      "size_bytes": 167168,
-      "crc32": "0x267d7f67",
+      "file": "tc-gu-01-slave-1.2.15.bin",
+      "version": "1.2.15",
+      "source_commit": "4d3a3b9",
+      "size_bytes": 167464,
+      "crc32": "0xbdbb2b4b",
       "sn_suffix": "s",
       "protocol": "V2.8 + 运动安全包络",
       "build": "local",
       "toolchain": "arm-none-eabi-gcc (15:13.2.rel1-2) 13.2.1 20231009",
-      "note": "Records the motor's own firmware version in follower flash, so it reads under MIT too (protocol V2.8): taken automatically on a private-protocol boot, or written by a host with the new 0x5C (Motor.set_motor_fw_version; a factory SOP step, and after a motor OTA). 0x58 returns the record when the motor cannot answer and says so in its former reserved byte. Model-dependent defaults now follow the motor model: auto-calibration stall torque (EL05 0.5 / RS00 1.8 N*m, was 0.35 for all -- too little to move an RS00, so calibration stopped short and could record the open direction backwards), default open direction (EL05 reverse, RS00 not), and a motion envelope enforced when none is stored (cont = stall rating, peak = rated torque -- 1.2.13; 1.2.12 used t_max, temperature wall 90/100 C; was no limit at all). When the model record changes, 0x700B, the calibration torque and the direction are reset to the new model's defaults; a stored calibration torque still at the old 0.35 migrates on boot. A gripper OTA no longer refuses when the motor is not on the bus. Carries 1.2.11 (control-UART receive restarted after an overrun). AFTER FLASHING, unplug the 24 V power cable for ~2 s and plug it back; the USB cable can stay in."
+      "note": "RS00 followers are handed: the gear train is mirror-mounted, so the open direction follows the side in the firmware SN (last digit odd = left = Reverse 1, even = right = Reverse 0) instead of one default per model. Auto-calibration uses that direction and corrects a stored one that disagrees (that run approaches slowly, since the old zero is at the open end); a model-record change picks the side's direction too. EL05 and units whose SN gives no side behave as 1.2.14. Verified on 0089s (left) and 0094s (right); 0089s deliberately stored reversed was corrected on the next 24 V power cycle. Carries 1.2.14 (motor firmware version record, protocol V2.8). AFTER FLASHING, unplug the 24 V power cable for ~2 s and plug it back; the USB cable can stay in."
     }
   }
 }

python/examples/ota_update.py

@@ -20,7 +20,7 @@
 of the update, not as troubleshooting.
 
 The released images ship in this repo under `firmware/`, and their filenames
-carry the version (tc-gu-01-slave-1.2.14.bin). Prefer the role selectors below:
+carry the version (tc-gu-01-slave-1.2.15.bin). Prefer the role selectors below:
 they read the current filename from firmware/manifest.json, so they keep
 working across releases, while a literal filename goes stale the next time the
 firmware is bumped. A name you do pass resolves against firmware/ from any
@@ -167,7 +167,7 @@ def _gripper_role(eps) -> str:
 def _default_firmware_for_role(role: str) -> str:
     """The shipped image for a role, named by firmware/manifest.json.
 
-    The filename carries the version (tc-gu-01-slave-1.2.14.bin), so it cannot be
+    The filename carries the version (tc-gu-01-slave-1.2.15.bin), so it cannot be
     derived from the role alone — the manifest is the single place the current
     release's filename is written, and `file` there is what this reads. That
     also means the manifest is load-bearing for every role-selector and --all
@@ -377,7 +377,7 @@ def _identify_image(fw_bytes: bytes):
     """CRC32 反查 firmware/manifest.json,返回 (role, meta) 或 (None, None)。
 
     **按内容认,不按文件名认。** 发布镜像的文件名带版本
-    (tc-gu-01-slave-1.2.14.bin),但文件名是可以被改的 —— 拷走、重命名、从别处
+    (tc-gu-01-slave-1.2.15.bin),但文件名是可以被改的 —— 拷走、重命名、从别处
     下载,名字就开始说谎。CRC32 不会:它认的是这一份字节。
 
     认不出来不是错误,只是"不是我们发布的镜像"(自己编的、第三方的),调用方按