Personal Research & Engineering for Embodied AI
This repository is my central hub for my robot learning.
Each project lives in its own orphan branch – completely independent, fully documented, and ready to run.
From simulation to real‑world, from VLA fine‑tuning to multi‑agent orchestration – this is where I build, break, and learn.
| Project | Branch | Tech Stack | Status |
|---|---|---|---|
| Running_SmolVLA | runnig smolvla |
PyTorch · LeRobot · LIBERO · Windows |
Complete |
| MRSLLM | multi-robot scheduler with LLM |
LLM · Disjunctive Graph · Harness · MRTA |
Complete |
| More coming soon... | – | – | 🔜 Planning |
Independent reproduction of Hugging Face's SmolVLA – trained and evaluated natively on Windows 11 (RTX 4060).
Overcame GPU OOM, path issues, and dependency hell to achieve 87%+ success rate on LIBERO tasks.
Closed‑loop system that combines LLM task planning, disjunctive graph scheduling, and simulation‑in‑the‑loop to solve MRTA‑Benchmark problems.
This repository uses orphan branches – each branch contains a self‑contained project with its own README.md, code, and data.
# Clone the entire repo (all branches)
git clone https://github.com/hy-0003/RobotLearning.git
cd RobotLearning
# Switch to a specific project branch
git checkout smolvla # for SmolVLA
git checkout MRSLLM # for multi‑robot schedulerI'm a freshman in Embodied AI.
Distributed under the MIT License. See LICENSE for more information.
Built by Yi He