$ cat ~/projects/action-models-book.md
Action Models for Robot Learning
#Book#Robotics#VLA#Action Models#Open Access#Vision-Language-Action
Open-access textbook on action models, vision-language-action systems, and modern robot learning — written and shipped as a static site with math, code highlighting, search, and AI-narrated audio.
A living, open-access textbook that distills the field of action models for robotics — from imitation learning fundamentals to vision-language-action (VLA) architectures. Built with Astro, KaTeX math, Shiki code blocks, Pagefind search, and per-section Kokoro-narrated audio.
// key_highlights
- ▸End-to-end arc from classical planning (STRIPS, PDDL, inverse kinematics) through deep RL and imitation learning to modern Vision-Language-Action foundation models
- ▸Hands-on Chapter 2 builds a working VLA from scratch using OpenVLA weights and the LIBERO simulator, walked through one line at a time
- ▸Dedicated chapters on the modern building blocks: sequence models for control, world models, and diffusion/flow-matching action heads
- ▸Deep dives on the foundation VLAs that define the field today — RT-1, RT-2, OpenVLA, Octo, π0, Helix, and GR00T N1 — with architecture, training data, and trade-offs
- ▸Practical Part 5 on fine-tuning a VLA for your own robot: dataset collection, LoRA vs. full fine-tuning, sim-to-real, evaluation, safety, and deployment