AM-ARM200

If you already have a working LeRobot environment and prefer to stay on the upstream repo, you can add STS3095 support manually with two changes.

Step 1 — Register STS3095 in the motor table

In lerobot/src/lerobot/motors/feetech/tables.py, add "sts3095" to each of the five model dictionaries. STS3095 is STS-series (same protocol as STS3215), so the values mirror that motor:

MODEL_CONTROL_TABLE = {
    ...
    "sts3095": STS_SMS_SERIES_CONTROL_TABLE,   # add this
}

MODEL_RESOLUTION = {
    ...
    "sts3095": 4096,                            # add this
}

MODEL_BAUDRATE_TABLE = {
    ...
    "sts3095": STS_SMS_SERIES_BAUDRATE_TABLE,   # add this
}

MODEL_ENCODING_TABLE = {
    ...
    "sts3095": STS_SMS_SERIES_ENCODINGS_TABLE,  # add this
}

MODEL_PROTOCOL = {
    ...
    "sts3095": 0,                               # add this (STS series = protocol 0)
}

Step 2 — Update the follower robot definition

In lerobot/src/lerobot/robots/so_follower/so_follower.py, replace the motor list with the AM-ARM200's 7-motor layout (adds wrist_yaw and swaps the three high-torque joints to sts3095):

motors=(
    ("shoulder_pan",  1, "sts3095", None),
    ("shoulder_lift", 2, "sts3095", None),
    ("elbow_flex",    3, "sts3095", None),
    ("wrist_flex",    4, "sts3215", None),
    ("wrist_yaw",     5, "sts3215", None),
    ("wrist_roll",    6, "sts3215", None),
    ("gripper",       7, "sts3215", MotorNormMode.RANGE_0_100),
),

After these two edits, the rest of the LeRobot workflow (calibration, teleop, recording, training) is identical to the standard SO-ARM100 setup.

Note: lerobot_alohamini already includes both changes — use the upstream path only if you have a specific reason to avoid the fork.