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

FeatureArduino UnoESP32Raspberry Pi 4Jetson Nano
OSNoneFreeRTOS / bareLinux (Debian)Ubuntu + JetPack
PythonNoMicroPythonFull CPython 3Full CPython 3
Wi-FiNoBuilt-inBuilt-inVia USB adapter
GPUNoneNoneVideoCore VI128-core NVIDIA
GPIO pins14 digital342840
Best forSimple I/OIoT + BLE + edge MLGeneral Linux, CVNeural inference

Hardware in the Lab

ComponentQuantity
Raspberry Pi 4 (4 GB)2
microSD cards (32 GB)4
Camera Module v2 (8 MP)2
Sense HAT1

There is currently no dedicated Raspberry Pi curriculum track at PAI Lab. For students interested in Linux-based physical computing, the recommended path is:

  1. Complete Arduino: Foundations of Physical Computing for hardware fundamentals
  2. Complete ESP32: IoT and Wireless Physical AI for networked sensing
  3. 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.