Inspire SDK - RH56 Robot Hand Controller

Python SDK for controlling the RH56 dexterous robot hand with optional wrist control.

This SDK provides two ways to control the hand:

• Hand Operator: High-level control using 3D fingertip target positions (inverse kinematics)
• Direct SDK: Low-level control using motor angles and percentages

Installation

pip install pyserial

Hand Operator (Inverse Kinematics)

The InspireHandOperator provides inverse kinematics to control the hand using fingertip target positions in 3D space. Simply provide where you want each fingertip to be, and the operator handles the conversion to motor commands.

Usage

from hand_operator import InspireHandOperator
import numpy as np

# Initialize operator (connects to hand automatically)
operator = InspireHandOperator(
    port='/dev/ttyUSB0',
    hand_id=1,
    forces=[40] * 6,
    speeds=[100] * 6
)

# Control hand using fingertip target positions (5 fingers × 3D coordinates)
# Input format: numpy array of shape (5, 3) with finger order: index, middle, ring, pinky, thumb
# Coordinates are in meters relative to hand base frame
target_positions = np.array([
    [0.1, 0.05, 0.02],  # index fingertip (x, y, z)
    [0.1, 0.03, 0.02],  # middle fingertip
    [0.1, 0.01, 0.02],  # ring fingertip
    [0.1, -0.01, 0.02], # pinky fingertip
    [0.08, 0.02, 0.03], # thumb fingertip
])

# Execute movement - returns motor control ratios
motor_angles = operator.step(target_positions)

# Optionally convert motor angles to joint angles for kinematics
joint_angles = operator.motor_to_joint(motor_angles)

# Reset hand to fully closed position
operator.reset()

Inverse Kinematics Visualization:

The green spheres represent target points, which are the recorded points plus noise. The operator then determines the motor percentage. Using this percentage, joint angles are calculated and provided to Mujoco.

https://raw.githubusercontent.com/NYU-robot-learning/inspire-sdk/main/img/ik_vis.mp4

Direct SDK Control

For direct control without inverse kinematics, use the SDK classes below.

Hand Only (RH56RobotHand)

from hand_sdk import RH56RobotHand

# Connect to hand
hand = RH56RobotHand(port='/dev/ttyUSB0', baudrate=115200, hand_id=1)
hand.connect()

# Set all fingers to 50% closed
hand.set_speeds([50] * 6)
hand.set_angles([50] * 6)

# Control individual finger
hand.set_finger(0, 75, speed=80)  # Finger 0 to 75% closed at 80% speed

# Read sensor data
angles = hand.get_angles()
forces = hand.get_forces()
state = hand.get_state()

# Quick actions
hand.open_hand(speed=50)
hand.close_hand(speed=50, force=50)

hand.disconnect()

Hand with Wrist (RH56RobotHandWithWrist)

from wrist_hand_sdk import RH56RobotHandWithWrist

# Connect to hand with wrist
hand = RH56RobotHandWithWrist(port='/dev/ttyUSB0', baudrate=115200, hand_id=1)
hand.connect()

# Control hand
hand.set_angles([50] * 6)
hand.set_speeds([50] * 6)

# Control wrist (pitch: -22.66° to 22.12°, yaw: -25.50° to 25.50°)
hand.set_wrist_angles(pitch=10.0, yaw=5.0, movement_time_ms=1000)
hand.center_wrist()

# Read wrist state
wrist_state = hand.get_wrist_state()
angles = hand.get_wrist_angles()  # Returns (pitch, yaw)

hand.disconnect()

Context Manager

with RH56RobotHand(port='/dev/ttyUSB0') as hand:
    hand.set_angles([100] * 6)
    time.sleep(2)
    hand.open_hand()

API Reference

RH56RobotHand

Connection:

• connect() -> bool - Open serial connection
• disconnect() - Close connection

Control:

• set_angles(angles: List[float]) - Set all 6 finger angles (0-100%, 0=open, 100=closed)
• set_speeds(speeds: List[float]) - Set movement speeds (0-100%)
• set_forces(forces: List[float]) - Set grip forces (0-100%)
• set_finger(index: int, angle: float, speed: float) - Control single finger

Reading:

• get_angles() -> List[float] - Get current finger angles
• get_forces() -> List[float] - Get current forces
• get_temperatures() -> List[int] - Get motor temperatures
• get_state() -> HandState - Get complete hand state

Convenience:

• open_hand(speed=50) - Fully open hand
• close_hand(speed=50, force=50) - Fully close hand
• clear_errors() - Clear error codes

RH56RobotHandWithWrist

Wrist Control:

• set_wrist_angles(pitch: float, yaw: float, movement_time_ms=1000) - Set both wrist angles
• set_wrist_pitch(pitch: float, movement_time_ms=1000) - Set pitch only
• set_wrist_yaw(yaw: float, movement_time_ms=1000) - Set yaw only
• center_wrist(movement_time_ms=1000) - Center wrist to (0°, 0°)

Wrist Reading:

• get_wrist_angles() -> Tuple[float, float] - Returns (pitch, yaw)
• get_wrist_state() -> WristState - Get complete wrist state

Joint Angles (for kinematics):

• set_joint_angles(joint_angles: List[float]) - Set joint space angles
• joint_angles_to_motor_angles(joint_angles: List[float]) - Convert joint to motor angles

GUI Applications

Tkinter GUI (hand only):

python hand_terminal.py

Web Interface (hand + wrist):

python wrist_hand_terminal.py
# Open http://localhost:5000 in browser