Lei Mao bio photo

Lei Mao

Machine Learning, Artificial Intelligence, Computer Science.

Twitter Facebook LinkedIn GitHub   G. Scholar E-Mail RSS


In the fall of 2017, I audited the “Self-driving Vehicles: Models and Algorithms for Autonomy (TTIC 31240)” course, which is informally called “Duckietown”, given by Prof. Matthew Walter at the Toyota Technological Institute at Chicago (TTIC). This is actually a new collaborative course that just starts from 2016. I first knew this course from the TTIC Duckietown website. After seeing some videos of autonomous Duckiebot vehicles made by students running in the Duckietown, I became deeply fascinated in this course.

In the beginning, I was eager to enroll in this course despite the fact that the course size is limited. However, I was prevented from enrollment by the department regulation because new students could only take at most three elementary courses in the first semester and the Duckietown course is considered an advanced course. I was not happy about this but later I found it was actually wise to audit the Duckietown course instead of actually take in my first semester at the University of Chicago because the course workload at the University of Chicago is tremendous. I barely just survived from the first semester where I took three elementary courses (One of the courses I took is “C Programming” in which I spend more than 35 hours weekly on weekly projects). But I wish I could formally attend this course next year when I totally get used to the study demand of the University of Chicago.

Course Content

The Duckietown course is actually a very comprehensive advanced PhD level course, although it is very open to master students and undergraduate students (I saw a few of them in the class). Unlikely the lovely duckie logo, the course content is mainly taught using mathematics and statistics. As a computer science student switched from biological science, I felt a lot of pressure from PhD classmates who seem to understand all the course contents orz. I should spend more time enhancing my background in mathematics and statistics in the near future.

Here I am going to talk about some of the contents that are most impressive to me.


Students are required to assembly their own autonomous dukiebot, install the system, and calibrate the mechanics of the robot. It is actually a very fun task to do if one has never done a similar job before. However, calibration requires some mechanics knowledge (math).

Computer Vision

The autonomous dukiebot uses a camera as its unique information source to guide its behaviors. The camera installed on the robot has to be calibrated according to some theories (math again). Prof. Walter also spent two to three lectures on modern computer vision techniques in feature detection, object recognition, etc. Although computer vision is a broad topic, the content in the course is very theoretical (math again).

Statistical Inference

Because the autonomous duckiebot is supposed to be intelligent, a lot of statistical inference techniques, such as particle filtering, are used in robot localization and decision making (here comes statistics). Several classical robotics problems, such as Simultaneous Localization and Mapping (SLAM), were also taught thoroughly in the class.


Because it is a relatively new course, many features of the dukiebot have not been standardized. Students from different backgrounds assembled in groups will work on a project to improve the functionality of the duckiebot (It could be software or hardware).

My Feelings

Although at first glance, you would be just playing with a robot and try to make it work as expected, the course is actually hard and the workload is also heavy. Unfortunately, I did not have time to review the course content after the class when I was auditing, but if I am going to really take it in the next year, I will expect I have to spend a non-trivial amount of time to understand all the theories behind. This course also reminded me that I am still to ignorant and I have to work harder to catch up with the PhD students and state-of-art researches. Prof. Walter is a very good professor.

Dukietown Family

The family includes Prof. Walter, TAs, PhD, master and undergraduate students from different backgrounds at the University of Chicago and TTIC (I am the one standing backmost). Thanks a lot for the Duckietown course.

The Duckietown Times