From gridlock to green lights: °µĶųTV student researcher uses AI to smooth traffic
āGrowing up in Dhaka, Bangladesh, Iāve been stuck in traffic a gazillion times,ā says Azizul Hakim ā26. āSo when the chance to spend my summer improving traffic lights came up, I jumped at it without second thoughts.ā
Azizul is a computer science major at °µĶųTV and a computer engineering student through the Collegeās 2-1-1-1 engineering program with Dartmouth College.
As part of °µĶųTV's Summer Faculty Student Research Program, he worked with Assistant
Professor of Computer Science Wenlu Du to . His simulations showed that AI-powered signals can significantly outperform traditional
systems ā offering a glimpse into a future where commutes are shorter and streets
flow more smoothly, both in the U.S. and back home.
A traffic simulation that Azizul used as part of his research.
Q: Can you give a quick summary of your research project?
We applied state-of-the-art deep reinforcement learning algorithms ā PPO, DQN, and
QL ā to optimize traffic signal timing at North Broadway in Saratoga Springs. Using
a traffic simulator, we generated realistic traffic demand and trained agents to control
signal phases. Our results showed that these reinforcement learned-based controllers
outperformed the current fixed-time traffic light settings, reducing delays and improving
flow.
Q: How does this project tie into your major or future goals?
As a computer science major, there are multiple career paths open to us after graduation. Machine learning is one of them. People often think of ChatGPT when they hear āmachine learningā these days, but machine learning research has been going on for years and has many other real-world applications. This project gave me a strong introduction to machine learning techniques and workflows ā skills that will be valuable whether I pursue a role in industry or continue to graduate study.
Q: What drew you to this research opportunity?
I grew up in Dhaka, Bangladesh, where severe congestion is often managed manually by traffic officers rather than by traffic signals. I was motivated to learn how intelligent traffic systems work here, with the hope of applying similar solutions back home someday.
Q: Whatās your working relationship been like with your faculty mentor?
Professor Du was very involved and supportive ā more of a mentor than a distant supervisor. She encouraged us to choose our own research questions and guided us throughout the process. Outside of work, she made meetings enjoyable; we often played Nintendo Switch or VR games and shared snacks.
Q: Whatās one really cool thing youāve learned or are still figuring out?
Iām fascinated by how neural networks and reinforcement learning algorithms can quickly learn efficient strategies for complex, real-world problems. The blend of theoretical design and practical performance in AI is especially impressive.
Q: What role does creativity or innovation play in your work?
Creativity was essential to building a realistic, safe simulation. I had to design signal-phase controls that were effective but not erratic, and ensure the system behaved reliably under varied traffic conditions. Balancing realism, safety, and performance required innovative thinking.
Q: How has this experience shaped your thinking about post-grad plans?
Before this project, I was set on going straight into industry. After this positive
research experience, Iām now open to graduate school if I find a compelling project
in a similar area.
Q: Whatās unique about doing this kind of project at °µĶųTV?
Machine learning is rapidly growing, and °µĶųTV now has faculty expertise in the area. That makes it an excellent place for computer science students to take machine learning courses and participate in hands-on research opportunities that have real-world impact.
Q: What else are you involved in at °µĶųTV outside of this project?
I will serve as co-president for both °µĶųTV Codes, a computer science-oriented club, and the Muslim Students Association this fall. Iām involved with Hayat, a club celebrating South Asian cultures, and the International Student Union. I will also be a peer mentor for a first-year Scribner Seminar taught by Associate Professor of Computer Science Aarathi Prasad.
