Robot Learning

Sebastian Thrun

Course Description

Robot learning is concerned with algorithms that enable robots to improve their performance with experience. Learning can make up for lack of knowledge when programming robots, such as lack of exact task specifications, lack of environment models, lack of sensor models, or lack of effective control strategies. Traditionally, the field of robotics has paid little attention to robot learning. With a new generation of intelligent service robots in close reach, and with the number of success stories increasing, learning is likely to become a fundamental part of robotics.

This tutorial will provide an introduction into the basic algorithms and techniques used in robot learning. It will cover recent work on learning models, learning control, and probabilistic reasoning. In addition, it will highlight some recent success-stories of robot learning and give some guidance for applying robot learning in practice.

The tutorial is targeted towards students, engineers, scientists, and teachers who are new to the field of robot learning, but who would like to get an overview of the field (and who would like to share my excitement).

Prerequisite Knowledge

Basic knowledge in robotics, machine learning, or statistics will be helpful but is not required.

About the Lecturers

Sebastian Thrun is a research faculty member at Carnegie Mellon University. His research interests lie in the areas of machine learning, neural networks and robotics. Thrun received his Ph.D. in 1995 and his M.Sc. in 1993, both from the University of Bonn in Germany. He is a consultant for several companies, including Real World Interface Inc., a leading US mobile robot manufacturer. Thrun recently co-edited a special issue of the journal Machine Learning and a book on "Robot Learning," and is currently editing another book entitled "Learning to learn."
Last modified: Thu Feb 20 14:07:12 JST 1997