A curated path through robotics — courses and books to learn the math, middleware and simulators
to build on, the perception / planning / control / manipulation stacks, and the learned-robotics
wave (RL, imitation, VLA foundation models). Opinionated and kept tight. Links open in a new tab.
Courses & learning
| Resource | What | Link |
|---|
| Modern Robotics (Lynch & Park, Northwestern) | The best free course + book on kinematics/dynamics/control. | site |
| Underactuated Robotics (Russ Tedrake, MIT) | Optimal control + dynamics for real robots. Full videos. | site |
| Robotic Manipulation (Tedrake, MIT) | Modern manipulation — perception to control. | site |
| ROS 2 Tutorials | Official path into the dominant robotics middleware. | docs |
| Probabilistic Robotics (Thrun et al.) | The reference for state estimation, SLAM, localization. | pdf |
Middleware & frameworks
| Resource | What | Link |
|---|
| ROS 2 | The de-facto robotics middleware — nodes, topics, services, DDS. | docs |
| MoveIt 2 | Motion planning + manipulation framework for ROS. | site |
| Nav2 | The ROS 2 autonomous navigation stack. | docs |
| ros2_control | Real-time hardware abstraction + controllers. | docs |
Simulators
| Resource | What | Link |
|---|
| Gazebo | The classic open ROS-integrated simulator. | site |
| MuJoCo | Fast, accurate contact physics — the control/RL default. | site |
| NVIDIA Isaac Sim / Isaac Lab | Photoreal, GPU-parallel sim for learned robotics. | site |
| PyBullet / Drake | PyBullet = easy RL sim; Drake = rigorous model-based control. | site |
Perception & state estimation
| Resource | What | Link |
|---|
| Open3D | Point clouds, registration, reconstruction. | site |
| PCL | The classic point cloud library. | site |
| ORB-SLAM3 / RTAB-Map | Visual / RGB-D SLAM systems. | repo |
| GTSAM / Ceres | Factor-graph + nonlinear optimization for estimation. | site |
Planning & control
| Resource | What | Link |
|---|
| OMPL | The open motion planning library (RRT*, PRM, etc.). | site |
| OCS2 / Crocoddyl | MPC + optimal control for legged/manipulator robots. | repo |
| Pinocchio | Fast rigid-body dynamics + derivatives. | repo |
| CasADi / acados | Nonlinear optimization + real-time MPC. | site |
Learned robotics & foundation models
| Resource / Paper | Why it matters | Link |
|---|
| LeRobot (HF) | Open library + datasets for real-world robot learning. Start here. | repo |
| RT-2 / RT-1 (Google) | Vision-language-action models — web knowledge into control. | arXiv |
| Open X-Embodiment / Octo | Cross-robot dataset + generalist policy. | site |
| Diffusion Policy (2023) | Visuomotor policies as conditional diffusion. Hugely influential. | arXiv |
| ALOHA / ACT | Low-cost bimanual teleop + imitation learning. | site |
| π0 (pi-zero) (Physical Intelligence) | Flow-based generalist VLA foundation policy. | post |
More curated lists
| Resource | What | Link |
|---|
| Awesome Robotics (Kiloreux) | Long-standing community link collection. | repo |
| Awesome Robotics Libraries | Curated index of robotics software/libraries. | repo |
| Awesome LLM Robotics | Papers on LLMs/VLMs for robotics. | repo |
where to start
New to robotics? Do Modern Robotics for the math, learn ROS 2 + Gazebo to build, then Tedrake's
Underactuated/Manipulation for control. For the learning wave: simulate in Isaac/MuJoCo, then
LeRobot + Diffusion Policy on real hardware. Sim-to-real gap is the eternal tax — budget for it.