← Back to projects
P-01 · Flagship

GRIP-5

FlagshipWorking PrototypeRoboticsEmbeddedAI Experiment

Verified summary

Gesture-responsive prosthetic-hand prototype mapping hand-tracking landmarks to five servo-driven finger states.

Problem or experiment question

Can webcam hand tracking drive a five-finger hardware prototype with stable discrete finger states?

What was actually built

  • Webcam frame input with MediaPipe hand tracking and 21 landmarks
  • Joint-angle calculation mapped to finger openness with EMA smoothing
  • Quantization into a five-character discrete state format
  • Serial or UDP transport to Arduino firmware
  • Five-servo proportional movement and rock-paper-scissors demo behavior

Architecture or workflow

  • Python capture and gesture logic on a host machine
  • Transport layer sends quantized finger states to MCU
  • Arduino parses packets and drives five servos

Important technical decisions

  • Discrete five-character states simplify MCU parsing over continuous angles
  • Exponential moving average reduces jitter before quantization
  • UDP/serial dual path supports lab prototyping flexibility
GRIP-5 signal path from webcam input through MediaPipe hand tracking with twenty-one landmarks, joint-angle geometry, finger openness mapping, exponential moving average smoothing, zero-to-nine quantization, and a five-character state string over serial or UDP transport to Arduino parsing and proportional easing on five servos. Zones group host processing, transport, microcontroller firmware, and servo output.SIG-GRIP · HAND-TO-SERVO PATHHOST · VISION & STATEWebcamframe inMediaPipehandsLandmarks21 ptsJoint anglesgeometryOpennessper fingerEMAsmooth0–9 QdiscreteFive-char statefinger codesTRANSPORTSerial / UDPfive-char packetMCUArduino parsevalidate & mapOUTPUTFive servoseased motionRPS demo pathverified pathdemo branch
System diagram

Public status

Working PrototypeImplementedExperimental
Python
MediaPipe
Arduino
Servo control
Serial/UDP