lerobot & SO 100

最新推荐文章于 2025-06-12 18:05:04 发布

原创最新推荐文章于 2025-06-12 18:05:04 发布 · 299 阅读

9 ·

CC 4.0 BY-SA版权

文章标签：

#lerebot #so100 #gr00t n1

一、概述

1、Lerobot：LeRobot 旨在为 PyTorch 中的真实机器人技术提供模型、数据集和工具

最顶层：scripts (命令行工具)，属于发号施令层
中间层：common/policies (控制策略)，属于中间执行层
底层支撑层：common/datasets, common/envs, common/robot_devices
基础配置层: configs, common/utils

2、So 100: 2个6关节的机械臂，主臂遥操作从臂

二、安装配置

Using the SO-100 with LeRobot

1、安装Lerobot & SO 100

conda create -y -n lerobot python=3.10
conda activate lerobot
git clone https://github.com/huggingface/lerobot.git ~/lerobot
conda install ffmpeg -c conda-forge
cd ~/lerobot && pip install -e ".[feetech]"

2、配置SO 100

找到与每个手臂关联的 USB 端口并更新配置文件

@RobotConfig.register_subclass("so100")
@dataclass
class So100RobotConfig(ManipulatorRobotConfig):
    calibration_dir: str = ".cache/calibration/so100"
    # `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
    # Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
    # the number of motors in your follower arms.
    max_relative_target: int | None = None

    leader_arms: dict[str, MotorsBusConfig] = field(
        default_factory=lambda: {
            "main": FeetechMotorsBusConfig(
                port="/dev/tty.usbmodem58760431091",  <-- UPDATE HERE
                motors={
                    # name: (index, model)
                    "shoulder_pan": [1, "sts3215"],
                    "shoulder_lift": [2, "sts3215"],
                    "elbow_flex": [3, "sts3215"],
                    "wrist_flex": [4, "sts3215"],
                    "wrist_roll": [5, "sts3215"],
                    "gripper": [6, "sts3215"],
                },
            ),
        }
    )

    follower_arms: dict[str, MotorsBusConfig] = field(
        default_factory=lambda: {
            "main": FeetechMotorsBusConfig(
                port="/dev/tty.usbmodem585A0076891",  <-- UPDATE HERE
                motors={
                    # name: (index, model)
                    "shoulder_pan": [1, "sts3215"],
                    "shoulder_lift": [2, "sts3215"],
                    "elbow_flex": [3, "sts3215"],
                    "wrist_flex": [4, "sts3215"],
                    "wrist_roll": [5, "sts3215"],
                    "gripper": [6, "sts3215"],
                },
            ),
        }
    )

配置12个舵机
手臂校准

三、使用

1、记录数据集

python lerobot/scripts/control_robot.py \
  --robot.type=so100 \
  --control.type=record \
  --control.fps=30 \
  --control.single_task="Grasp a lego block and put it in the bin." \
  --control.repo_id=lerobot/so100_test \
  --control.tags='["so100","tutorial"]' \
  --control.warmup_time_s=5 \
  --control.episode_time_s=30 \
  --control.reset_time_s=30 \
  --control.num_episodes=2 \
  --control.push_to_hub=false
  # 在此数据上继续采集
  --control.resume=true

核心参数
- repo_id：数据集名称，默认路径 ~/.cache/huggingface/lerobot；可通过环境变量HF_LEROBOT_HOME修改
- reset_time_s：录制一集的最大时长
- reset_time_s：场景重置时间
- num_episodes：录制多少集
录制技巧
- -> 提前退出，保存当前集，并重置环境
- <- 提前退出，丢弃当前集，并重置环境
- eac 保存当前集，并结束采集
合并多个数据集
- Improved Dataset Merge Script for Multiple Dataset Processing

2、可视化数据集

python lerobot/scripts/visualize_dataset_html.py \
  --repo-id lerobot/so100_test \
  --local-files-only 1

3、重播

python lerobot/scripts/control_robot.py \
  --robot.type=so100 \
  --control.type=replay \
  --control.fps=30 \
  --control.repo_id=lerobot/so100_test \
  --control.episode=0

4、训练

python lerobot/scripts/train.py \
  --dataset.repo_id=lerobot/so100_test \
  --policy.type=act \
  --output_dir=outputs/train/act_so100_test \
  --job_name=act_so100_test \
  --policy.device=cuda \
  --wandb.enable=true

5、推理

python lerobot/scripts/control_robot.py \
  --robot.type=so100 \
  --control.type=record \
  --control.fps=30 \
  --control.single_task="Grasp a lego block and put it in the bin." \
  --control.repo_id=lerobot/eval_act_so100_test \
  --control.tags='["tutorial"]' \
  --control.warmup_time_s=5 \
  --control.episode_time_s=30 \
  --control.reset_time_s=30 \
  --control.num_episodes=10 \
  --control.push_to_hub=true \
  --control.policy.path=outputs/train/act_so100_test/checkpoints/last/pretrained_model