NVIDIA Jetson Nano at PAI Lab

PAI Lab의 엣지 AI 및 로봇공학 연구에서 Jetson Nano의 역할.

NVIDIA Jetson Nano at PAI Lab

The Jetson Nano is planned as the next step in the PAI Lab hardware stack — the first platform in our curriculum that bridges from microcontrollers to full GPU-accelerated edge AI inference. Where Arduino and ESP32 handle real-time I/O, the Jetson Nano runs full neural networks on-device.

Why Jetson Nano?

• GPU compute — 128-core NVIDIA Maxwell GPU runs YOLO, segmentation, and depth models at real-time rates
• Full Linux — Ubuntu 18/20 via JetPack SDK; Python, apt packages, ROS 2, and SSH access out of the box
• NVIDIA ecosystem — TensorRT, CUDA, cuDNN, DeepStream — the same tools used in production edge AI
• Physical AI target — run perception models alongside ROS 2 nodes on the same board
• Natural progression — Arduino → ESP32 → Jetson Nano is a deliberate complexity ramp in the PAI curriculum

Hardware Status

Planned — The PAI Lab plans to acquire Jetson Nano units. Lab exercises and curriculum for this platform will be developed once hardware is in place.

Planned Curriculum

• Jetson Nano: GPU-Accelerated Physical AI — a planned 8-unit track from initial JetPack setup through real-time neural inference on robot platforms

Use Cases in PAI Lab Research

Once hardware is available, the Jetson Nano is planned for:

• Real-time object detection (YOLO) on mobile robot platforms
• On-device TensorRT model optimization and inference benchmarking
• ROS 2 perception nodes running alongside sensor fusion pipelines
• Side-by-side benchmarks with ESP32 TFLite micro models — same task, different hardware
• Student research projects requiring GPU compute without cloud dependence

Jetson Nano vs. Other Lab Platforms