CS 4756 - Robot Learning

Spring. 4 credits. Letter grades only.

Prerequisite: MATH 1920 or MATH 2220, MATH 2940, CS 1110, and CS 4780 or permission of instructor. Co-meets with CS 5756.

S. Choudhury.

Advances in machine learning have proved critical for robots that continually interact with humans and their environments. Robots must solve the problem of both perception and decision making, i.e., sense the world using different modalities and act in the world by reasoning over decisions and their consequences. Learning plays a key role in how we model both sensing and acting. This course covers various modern robot learning concepts and how to apply them to solve real-world problems.

Outcome 1: Learning perception models using probabilistic inference and 2D/3D deep learning.

Outcome 2: Imitation and interactive no-regret learning that handle distribution shifts, exploration/exploitation.

Outcome 3: Practical reinforcement learning leveraging both model predictive control and model-free methods.

Outcome 4: Open challenges in visuomotor skill learning, forecasting and offline reinforcement learning.