Monday, October 8, 2018

OSCAR Student Zhanpeng Chen Turns Ambiguous Traffic Scenario Into Autonomous Vehicle's Intelligence

­­­­My research project is Turning Ambiguous Traffic Scenario Into Autonomous Vehicle’s Intelligence. The goal of this OSCAR project is to develop a database with annotated driving videos that can be used as a training set for artificial neural networks in future autonomous vehicles. We have been seeing many successful examples of neural networks in recent years; however, a large amount of data is required in order to successfully train a useful network. This project will serve as a stepping-stone for future autonomous car projects.

I’m working with 5 other undergraduate students as well as two high school students. This is an incredibly multidisciplinary project; we came from different backgrounds: computer science, psychology, and physics. We are collectively solving the same problem using different knowledge. A lot of research had been done in this area; however, there are only a few researches that were done with multiple disciplines. We are not using just machine learning techniques to solve this problem; we also looked at this problem from a human factors perspective. I have been working with these wonderful people. I learned a lot about discipline other than mine from these people.


I meet with other students in this project on a daily basis. We also receive mentorship from the project advisors on a weekly basis. I typically spend most of my time in front of a laptop writing code and working with the video data. I personally focused on lane detection for autonomous vehicles to detect drivable areas. Most of my work had been done using computer vision techniques. I also read many research articles to gain more knowledge of this topic. We do daily stand-ups and weekly report to share our work as well as to gain a better understanding of other people’s effort. Overall, this research project has been a truly enriching experience for my undergraduate study.