Jianyu Su, a native of Hubei, China; is an avid bodybuilder in his free time and also enjoys being amongst friends and real world problem solving. He is pursuing a PhD in the University of Virginia School of Engineering's Systems Engineering program and works with Professor Peter Beling at the Link Lab on projects related to artificial intelligence (AI) for multi-agent systems.What three words have described your life in the last year?
Boring, but fruitful.
Why did you choose those words to describe your life in the last year?
I enjoyed my life in the Link Lab prior to the pandemic. I can socialize with people on a daily basis, which, at this time, is impossible. In addition, I went to the gym with my Chinese gang regularly and I have been working out at AFC for so long that many people in the gym know me. I have skipped the gym for five months straight since the pandemic started, therefore, my life is boring. However, I have written several papers since February, with two papers accepted, the others under review. I lead the writing of a CCI proposal with my advisor Professor Peter A. Beling and we are pushing an AI traffic testbed with researchers from VT. Most importantly, I finished my dissertation proposal this May. To this extent, I think it has been fruitful.
What is something you wish other people knew about you?
I wish people would know me as a competitor, a good friend, and maybe a bodybuilder.
Why is it important to you that people know you as “a competitor”?
I think many people might know me as a funny guy. However, it is the competitor in me that drives me to work out, to keep my everything in shape. I am privileged to have some friends, who not only constantly challenge me on heavy weights in the gym but also inspire me to conduct thorough research. I hope I imposed the same effect on them as they did on me. In addition, it is important to have this competitor mindset in the PhD business. I have to constantly think about how to contribute to the current literature.
What kind of projects have you worked on while at the Link Lab?
I work on artificial intelligence for multi-agent systems. For example, I have a project to predict an ego vehicle's actions while taking into consideration the ego vehicle's neighbor vehicle's actions. My coolest project is to build a multi-agent AI for StarCraft II games. Reinforcement learning emerged recently and AI typically beats humans on many games such as Go, Dota2, and StarCraft. StarCraft differs from other games in that there are many units in the game, which makes it very challenging. My previous research focused on utilizing communication to achieve cooperation between agents. My current research, which aims to propose a general-purpose multi-agent AI that does not necessarily rely on communication, benchmarks our newly proposed AI with other state-of-the-art AIs on the StarCraft II micromanagement games. You can find the videos of my AI winning the game on my GitHub page. It is interesting that the AI-controlled units actually learned good policies.
Humans of Link Lab is a new initiative of UVA's Link Lab to exhibit the interesting people who work in the space and their passions both inside and outside of the Lab. To see more content, connect with us on Instagram (@linklabatuva), Twitter (@UVALinkLab), and LinkedIn (UVA Engineering Link Lab).