From classical kinematics to VLAs — the complete robot learning stack with runnable LeRobot code.
Why robot learning matters, the LeRobot ecosystem, datasets (LeRobotDataset), streaming, teleoperation, and data collection.
Forward/inverse kinematics, Jacobians, differential IK, feedback control. Why dynamics-based methods hit their limits.
MDPs, Q-learning, DQN, DDPG, SAC, domain randomization, HIL-SERL. Learning from trial and error on real hardware.
VAEs, diffusion models, flow matching, ACT, Diffusion Policy, temporal ensembling, async inference. Learning from demonstrations.
VLAs, cross-embodiment transfer, π0, SmolVLA, and the path to foundation models that act across tasks and robots.