SuperBot

SuperBot is a modern, modular robotics framework designed for controlling dual-arm piper robots and other robotic systems. It provides a clean, intuitive API for robot control, sensor integration, and utility functions.

Features

• Modular architecture with separate modules for bots, sensors, and utilities
• Dual-arm piper robot control interface
• Sensor management system
• Utility functions for robotics operations
• Clean, well-documented API

Installation

pip install superbot
# Touch + Wuji dexterous hands (Canonical-70 embodiment):
pip install -e ".[wuji]"

DualTouch + Wuji (Canonical-70)

Unified embodiment for bimanual Touch arms + Wuji hands. Policy I/O uses the Canonical-70 layout (same as N3 dataset /lumos-vePFS/suda/Hopper/dataset/id_101_0825_canonical70_gr00t/meta/modality.json).

from superbot.bots.dual_touch_wuji import DualTouchWujiRobot

robot = DualTouchWujiRobot(can_interfaces=("can1", "can0"), enable_hands=True)
state70 = robot.get_state70()          # (70,) ABS EEF rot6d + Can22 hands
robot.execute_action70(action70)       # absolute 70-D command
robot.smooth_go_home()
robot.close()

Helpers live on superbot.utils.canonical70.C70 (pack_state70, action70_to_target, legacy_actions52_to_action70, …). Hand driver is wujihandpy only (superbot.bots.wuji_hand_replay.WujiHandReplayController).

Full Wuji feature map and data-conversion CLI: docs/features_wuji.md.

Dataset conversion (Dexora 39→70, XR→70): pip install -e ".[conversion]" then python -m superbot.conversion.convert_lerobot_to_canonical70 / python -m superbot.conversion.xr_to_lerobot_canonical70.

OpenCV 摄像头预览（Linux）

如果你只想快速查看 USB 摄像头画面，可以直接使用仓库里的 see_umi.py。

1) 安装依赖

pip install opencv-python

2) 查看可用摄像头编号

python find_opencv.py

输出会显示可用摄像头编号（如 0、1、2）。

3) 启动预览

默认打开 source=0：

python see_umi.py

指定摄像头编号：

python see_umi.py --source 1

推荐参数（兼容性较好）：

python see_umi.py --source 0 --format MJPG --backend v4l2 --width 640 --height 480 --fps 30

退出：按 q 或 Esc。

4) 常用参数

• --source：摄像头设备号（默认 0）
• --format：像素格式，支持 AUTO / MJPG / YUYV
• --backend：采集后端，支持 v4l2 / auto
• --width --height：目标分辨率
• --fps：目标帧率

5) 画面异常排查

当出现花屏、偏色、卡顿时，通常是像素格式或分辨率不匹配：

1. 优先尝试 MJPG：

   python see_umi.py --source 0 --format MJPG --backend v4l2 --width 640 --height 480
   

2. 若仍异常，改为 YUYV：

   python see_umi.py --source 0 --format YUYV --backend v4l2 --width 640 --height 480
   

3. 观察程序启动时打印的实际参数（format/size/fps），确认驱动是否应用了你的设置。

Quick Start

from superbot import RobotWrapper

# Initialize the robot wrapper with CAN interfaces
robot = RobotWrapper(can_interfaces=("can_left", "can_right"))

# Move to a specific pose
import numpy as np
target_pose = np.eye(4)  # 4x4 identity matrix as example
robot.move_to_pose(left_target=target_pose, right_target=target_pose)

# Close the grippers halfway
robot.set_gripper(0.5, 0.5)

# Get current robot state
state = robot.get_current_state()
print(state)

Architecture

SuperBot follows a modular design:

• bots: Contains robot-specific implementations (dual_piper, etc.)
• sensors: Sensor management and interfaces
• utils: Common utility functions for robotics operations
• transport: Zero-copy instant VLA data transportation lib powered by memap && zeromq, fastest solution to sending data to poliy server.

Zero Transport

zero transport uses zmq and mmap for zero-copy data sharing, unlike most VLA policy server using http sending msgpack or binary, we are currently support data types contains:

• array;
• string (scalars);

Remote policy servers also expose a quick connectivity check by default:

curl http://<server-host>:8080/echo
# hello world, superbot

Contributing

Contributions are welcome! Please see the contributing guidelines for details.

报错记录

load extensions from /home/lumos7/miniconda3/envs/startouch/lib/python3.10/site-packages/extensions 2026-03-23 19:14:56.429 | ERROR | superbot.bots.dual_touch:_import_single_arm_right:107 - 无法导入 SingleArm，请确认 /home/lumos7/startouch-v1/interface_py 下存在 startouchclass.py: libstartouch.so: cannot open shared object file: No such file or directory client.sh: 行 12: --instruction: 未找到命令

修复：

编辑库配置文件

sudo nano /etc/ld.so.conf.d/startouch.conf

在文件中添加一行库文件所在的目录路径（替换成你实际的路径）

/home/lumos7/startouch-v1/lib

保存退出（按 Ctrl+O 回车，再按 Ctrl+X）

更新系统库缓存

sudo ldconfig

验证（能看到 libstartouch.so 即可）

ldconfig -p | grep startouch