Learning for Robotics

Robots have to perceive a messy world, decide what to do, and actually move — this course is about the modern toolkit for all three, blending classical robotics (kinematics, motion planning, control) with the deep learning and reinforcement learning methods that now drive perception and policy. You'll work through four homeworks, five quizzes, a midterm, and a sizable term project that pulls the pieces together, typically involving simulated manipulation. It sits at the graduate ML/robotics intersection, so comfort with linear algebra, probability, and Python-level deep learning is assumed; expect it to feed directly into thesis work on manipulation, VLAs, or learned control.