PAI Lab에서 Linux 기반 피지컬 컴퓨팅 및 프로토타이핑 플랫폼으로서의 라즈베리 파이.
Raspberry Pi at PAI Lab
The Raspberry Pi occupies a different niche from both Arduino/ESP32 (microcontrollers) and the Jetson Nano (GPU-accelerated inference). It is a full Linux single-board computer — ideal for projects that need a flexible developer environment, rich networking, and broad library support without specialized GPU hardware.
Why Raspberry Pi?
- Full Linux — Raspberry Pi OS (Debian-based); Python, apt, SSH, git, and the entire Linux ecosystem
- GPIO compatibility — 40-pin header bridges the gap between microcontrollers and a full OS
- Networking — built-in Wi-Fi and Ethernet; ideal for IoT dashboards, MQTT brokers, and networked sensing
- Camera support — CSI camera connector, PiCamera library, OpenCV integration
- Community — the largest single-board computing community; answers exist for virtually every problem
- Affordable — Pi 4 (8 GB) and Pi 5 offer substantial compute at low cost
Raspberry Pi vs. Other Lab Platforms
| Feature | Arduino Uno | ESP32 | Raspberry Pi 4 | Jetson Nano |
|---|---|---|---|---|
| OS | None | FreeRTOS / bare | Linux (Debian) | Ubuntu + JetPack |
| Python | No | MicroPython | Full CPython 3 | Full CPython 3 |
| Wi-Fi | No | Built-in | Built-in | Via USB adapter |
| GPU | None | None | VideoCore VI | 128-core NVIDIA |
| GPIO pins | 14 digital | 34 | 28 | 40 |
| Best for | Simple I/O | IoT + BLE + edge ML | General Linux, CV | Neural inference |
Hardware in the Lab
| Component | Quantity |
|---|---|
| Raspberry Pi 4 (4 GB) | 2 |
| microSD cards (32 GB) | 4 |
| Camera Module v2 (8 MP) | 2 |
| Sense HAT | 1 |
Related Curriculum
There is currently no dedicated Raspberry Pi curriculum track at PAI Lab. For students interested in Linux-based physical computing, the recommended path is:
- Complete Arduino: Foundations of Physical Computing for hardware fundamentals
- Complete ESP32: IoT and Wireless Physical AI for networked sensing
- Then work with Raspberry Pi independently — Python GPIO (
RPi.GPIO,gpiozero), OpenCV, and Flask/FastAPI for sensor dashboards are all direct extensions of ESP32 IoT skills
Raspberry Pi for Edge AI
The Raspberry Pi can run lightweight edge AI models via:
- TFLite — TensorFlow Lite runs on the Pi CPU; suitable for small models
- Coral USB Accelerator — Edge TPU module adds hardware ML inference via USB
- OpenCV DNN — MobileNet and similar models run well on Pi 4/5 with OpenCV
For models that require a GPU, the Jetson Nano is the next step.