COMP 766 / ECSE 683

Learning and Optimization for Robot Control

Course Description:

In this course, we will introduce concepts from a broad spectrum of motion control and motion planning approaches involving optimization and machine learning. The course materials consist of both lecture notes and seminal papers from the robotics literature. The specific robotic systems in which we will focus are manipulators and legged systems.

The class will begin with lectures on the fundamentals of recent research topics. Throughout the semester, each student will work on their chosen research projects that they present at the end of the semester.

Prerequisites

This class is designed for graduate students with a basic understanding of robot kinematics/dynamics, optimization, and machine learning. Experience with one of the robotic simulators is highly recommended. If you are not sure, feel free to contact me.

Marking Scheme (Tentative):

• Assignments (30%)

• Research paper presentations (30%)

  • Literature Presentation (10%)

  • Proposal Presentation (10%)

  • Final Presentation (10%)

• Project (40%)

  • Proposal (5%)

  • Literature review (5%)

  • First D

  • Peer review (10%)

  • Baseline implementations (10%)

  • New implementations (5%)

  • Final paper (20%)

Topics of interest (tentative):

• • Trajectory optimization, mixed-integer optimization, linear complementarity constraints

  • Imitation learning

  • Legged robots

  • Recent topics: transfer learning, meta-learning, etc