16-899D Statistical Techniques in Robotics is a graduate-level course that explores mobile robotics from a statistical perspective. In recent years, statistical techniques have changed the face of mobile robotics in many ways, providing robust new solutions to hard robotic problems, involving sensor uncertainty. Apart from being robust in practice, these statistical techniques also have a sound mathematical basis that makes it easy to understand the virtues and limitations of these approaches.

The course should be of interest to anyone seeking to develop robust robot software, and anyone who is interested in real-world applications of statistical theory. Students participating in this course will acquire the skill of developing robust software for robots operating in real-world environments, and understanding the mathematical underpinnings of their software. Even though this course focuses on mobile robotics, the techniques covered in this course apply to a much brooder range of embedded computer systems, equipped with sensor and actuators.

The course involves three types of activities:

  • Interactive classroom sessions, where students together with the instructor explore the basic mathematical foundations behind a range of popular robotics algorithms. Some of the sessions will take the form of traditional-style teaching, whereas others will be dedicated to brainstorming on challenging open problems.
  • Homework assignments will provide an opportunity to deepen the problem solving skills acquired in class.
  • Robot programming assignments will enable students to develop practical robot software, while deepening their understanding of the relation of mathematical calculus and the "real world."

Statistical techniques in roboticd is a new course which has not been offered at Carnegie Mellon University before. It has previously been taught at Stanford University, where it led to a number of publications at top conferences in the field. Detailed information on the Stanford course can be found here.

Prerequisites: This is a graduate level course tha tis pretty much self-contained. Familiarity with basic statistical concepts (Bayes rule, PDFs, Kalman filters, continuous distributions...) will be helpful for this course, as will be hands-on experience with software development in C or C++. But the most important prerequisite will be creativity and enthusiasm.