HRI
MiMo Mini — interactive hexapod robot
A bilingual EN/JA hexapod that detects a hand wave, recognises faces, and responds to voice commands on a Raspberry Pi 5.
PythonMediaPipeInsightFaceespeak-ngRaspberry Pi 5ROS 2MkDocs
Problem
Make a small legged robot socially interactive on low-cost edge hardware, in both English and Japanese, without cloud dependencies.
Approach
- Hand-wave detection with MediaPipe and face recognition with InsightFace.
- Bilingual English/Japanese voice command handling with on-device speech recognition.
- Bilingual TTS via espeak-ng using kana to avoid kanji mispronunciation.
- Modular Python architecture separating the vision module from application logic.
Result
- Runs on a Raspberry Pi 5 with camera, microphone, and speaker.
- Published a bilingual EN/JA documentation site (MkDocs Material, i18n) covering assembly, inverse kinematics, gaits, and ROS 2 integration.
System breakdown
Hardware
- Hexapod chassis with 18 servo joints
- Raspberry Pi 5
- Camera, microphone, and speaker peripherals
Firmware / embedded
- Servo gait sequencing and inverse kinematics
- Audio I/O device configuration on Pi 5
Software
- MediaPipe gesture pipeline, InsightFace recognition
- Speech recognition + espeak-ng kana TTS
- Modular Python application layer
- MkDocs Material bilingual documentation site
Project demo video slot