Cluster 11

Fundamentals of Machine Learning in Robotics

Instructors:
Ricardo Sanfelice, PhD
UCSC Department of Electrical and Computer Engineering
Jason Eshraghian, PhD
UCSC Department of Computer Science and Engineering

Prerequisite: none

Summary: This cluster will introduce the students to the basic principles of machine learning as applied to robotics and feedback control. This cluster will also introduce machine learning, models of robots, feedback control algorithms to accomplish a given goal, and methods to implement and validate a robotic system in a simulation environment.

All students in this cluster will be enrolled in the following courses:

Machine Learning for Robotic Control Applications

Machine learning is rapidly changing how we interact with the world. Training large-scale machine learning models is incredibly expensive. For example, the power bill of training ChatGPT exceeded millions of dollars. There are many applications where this amount of energy is not available, such as in portable robotics. This course will introduce the basics of machine learning, neural networks, and reinforcement learning to accomplish robotic control tasks. Emphasis will be placed on low-power machine learning techniques. The lectures will focus on training and using neural networks in Python and using reinforcement learning for robotic control. At the completion of this course, students will be able to define and apply fundamental concepts in machine learning to robotics, programming neural networks to perform image classification, and simulate the entire system in Python. 

Robotic Control Applications

Feedback control is the science that enables most engineering systems of today.  Robotic systems use algorithms that rely on real-time information to make decisions that affect the motion of the components defining robots.  This course will introduce models of robots, control algorithms to accomplish a given goal, and methods to implement and validate a robotic system in a simulation environment. The lectures will focus on mathematical modeling, design of algorithms, and computer simulation in Matlab and Python.  We will employ a ground vehicle as the driving robotic system on which the concepts and ideas are illustrated.  At the completion of this course, the students will be capable of translating specifications into properties of a robotic system, formulate basic mathematical models, employ and tune feedback and machine learning algorithms in the literature that are suitable to meet the given specifications, and simulate in Matlab robotic systems.