The library
Everything on the path, and beyond it
Courses, books, simulators, code, datasets, hardware and the people worth following. Each one is tagged with the month where it fits.
38 resources
Stephen Boyd, Stanford
54 short lectures, ~18.5 h. Linear algebra as a tool, not proofs.
CoursesM1Oussama Khatib, Stanford
16 lectures, ~19 h. Kinematics, the Jacobian, dynamics, control.
CoursesM2Fei-Fei Li and team, Stanford
18 lectures, ~21 h. Ends on robot learning.
CoursesM3Stanford
IMUs, lidar, cameras, and why real sensors drift.
CoursesM3Marco Pavone and Daniele Gammelli, Stanford
19 lectures, Spring 2026. LQR, MPC, then RL.
CoursesM4Chelsea Finn, Stanford
18 lectures, ~20 h. Lecture 16 is RL for real robots.
CoursesM5Emma Brunskill, Stanford
Slower and more mathematical, if CS224R moves too fast.
CoursesM5Andrej Karpathy
10 videos, ~19 h. From one neuron to GPT-2, every line typed live.
CoursesM3Sergey Levine, UC Berkeley
25 lectures + homework code, ~36 h. Spring 2026.
CoursesM5Russ Tedrake, MIT
23 lectures, ~30 h. Grasping, planning, behavior cloning. Exercises on Drake.
CoursesM6Russ Tedrake, MIT
21 lectures, ~28 h. The math behind robots that walk, run and flip.
CoursesM4Stanford ASL
New talks every quarter. One a week keeps you current.
CoursesM6Boyd & Vandenberghe
Free PDF with exercises. The ENGR108 companion.
BooksM1Gammelli, Lorenzetti, Luo, Zardini, Pavone
Free online, Python exercises in Jupyter. The AA203 companion.
BooksM4Russ Tedrake
Free website, updated every semester.
BooksM6Russ Tedrake
21 chapters, every example runs as a notebook.
BooksM4Google DeepMind
The main simulator on this path. Runs on a normal laptop, no GPU.
SimulatorsM2Google DeepMind
Tuned models of real robots: arms, humanoids, robot dogs.
SimulatorsM2Hugging Face
Simulated tasks, datasets and policies. The same code drives a real SO-101.
SimulatorsM6NVIDIA
The industry tool for humanoid RL. Needs an NVIDIA RTX GPU.
SimulatorsMIT / Toyota Research
Behind the MIT manipulation and underactuated courses.
SimulatorsFarama
Cart-pole, Hopper and friends, ready for control and RL.
SimulatorsM4DLR-RM
Ready PPO and SAC for your first RL runs.
CodeM5Physical Intelligence
The π0 and π0.5 robot brains, open weights.
CodeM6community
200+ robot files you can load in one line of Python.
CodeM2rerun.io
Camera, lidar, transforms and joints on one timeline for debugging.
CodeGoogle DeepMind and partners
Data from 22 different robots in one dataset.
DatasetsM6Cheng Chi et al.
Robot moves generated the way AI generates images.
PapersM6- SO-101 arm ↗$122
TheRobotStudio / Hugging Face
One open-source follower arm, priced from the official parts list. $230 for the leader + follower pair.
HardwareM6 - Reachy Mini ↗$399
Pollen Robotics
A desktop robot from Hugging Face. $499 for the wireless version.
Hardware - Go2 Air ↗$1,600
Unitree
Robot dog launch price. Resellers ask around $2,400.
Hardware - HopeJR ↗~$3,000
Hugging Face
An open-source humanoid, Hugging Face's own estimate per unit.
Hardware - G1 ↗$13,500
Unitree
The humanoid in most viral demos, base model.
Hardware @chelseabfinn
Co-founder of Physical Intelligence, teaches CS224R.
People@svlevine
Co-founder of Physical Intelligence, teaches Berkeley CS 285.
People@DrJimFan
NVIDIA's Director of Robotics, co-leads the GR00T lab.
People@YunzhuLiYZ
Taught the robot learning lecture in CS231N.
People@pabbeel
Berkeley professor, deep RL for robots.
People