Robotics is an important area, with a range of applications from industrial automation to healthcare and assistive technologies. An important problem in robotics has been building agents that are able to operate in complex, unstructured human-centric environments like homes, hospitals and offices. Machine learning provides a potential solution to build these types of adaptive robotic systems. In particular, deep learning has shown the ability to learn complex predictive behaviors from large amounts of high dimensional data like images or text, with minimal amounts of hand-design. The question remains - how can these deep learning methods be useful in robotics?

In this course, we examine how we can leverage deep learning methods to build robotic learning systems that can adapt and continue improving in real world applications. This course aims to provide an understanding of how deep learning methods can be useful for robotics, with an in depth look into research frontiers and representative papers. We plan to cover a wide range of methods: reinforcement learning in model-based and model-free settings, imitation learning and offline reinforcement learning, multi-task and meta learning, transfer learning in robotics and many more topics at the frontier of robotic learning.

The expected course outcomes are:

  • Understand the challenges and advantages of using learning methods for robotics problems
  • Gain perspective on the fundamental techniques behind advances in robot learning
  • Learn how to model robotics problems through the framework of reinforcement learning across various applications
  • Gain hands-on experience with implementing and tuning learning algorithms for robotics


Undergrad and master students, please use this form to apply for an add code

Lecture: MW 1:30 - 2:50pm, CSE2 G04
Instructor: Abhishek Gupta (abhgupta at cs); Office Hours: F 4 - 5pm Gates 215
TA: Liyiming Ke “Kay” (kayke at cs); Office Hours: M 3 - 4pm Gates 153
We use Ed for announcement/QA and Canvas for assignment submission


The course assumes a strong footing in the following, comparable to taking a graduate or advanced undergraduate level course:

  • Machine learning
  • Linear Algebra
  • Calculus
  • Probability theory
  • Proficiency in python

Homework and Grading

The content here is tentative and subject to change

  • Final Project (50%): The course project is the principal outcome of this course.
  • Paper Discussions (35%): Student-led discussions on frontier research papers.
  • Paper Commentary (15%): Typeset comments on the assigned readings.
  • Homework (Optional, 7%): Light-weight projects to get you started with some algorithms or make up for missing attendence in the paper discussion.


We thank Sergey Levine and Chelsea Finn for helping with course materials, etc.
We thank Brian Hou for creating and sharing the course website template.