Physical AI Paper Deep-Dive & Hands-on Deployment
Explore cutting-edge Physical AI research, break down key models and algorithms with Seeed robotic arms
A course that will always be free, carefully developed by the SeeedStudio AI Robotics team, paired with the 100% open-source reBot robotic arm, available for commercial use and reproduction. Complete it and become the next robotics expert!
- Always free
- 100% open source · Commercial use
- Hands-on · Reproducible
Hands-on Deployment
Start with platform preparation and data collection, follow the steps to train and deploy ACT, then learn to switch from ACT to other policies and gradually master the reproduction workflow for different algorithms.
Paper Deep-Dives
Read papers by topic, from basic imitation learning policies to vision-language-action models.
Part 1 · Action Chunking & Multi-Task Policies
Part 2 · Lightweight VLA Models & Efficient Deployment
Part 3 · π Robot Foundation Models
- π₀ — General-Purpose Robot Control
- π₀-FAST — Efficient Action Tokenization
- π₀.₅ — Open-World Generalization
Part 4 · Generalist VLA & Cross-Embodiment Adaptation
- NVIDIA GR00T — Adaptation & Fine-Tuning
- EO-1 — Interleaved Multimodal Pre-Training
- WALL-OSS — Embodied Reasoning & Robot Control
Part 5 · World Models & Video-Action Learning
Ongoing Updates
Part 6 · Unified Evaluation & Research Summary
Ongoing Updates