Research

At 30,000 feet, my PhD research aims to make RL deployable and actually deployed in roadway intelligent transportation systems to improve traffic efficiency, safety, sustainability, etc. From another perspective, I strive to bridge the gap between physical systems (roadway infrastructure, vehicles) and computational processes (analysis, observation, control, and optimization) for physical systems, a concept known as cyber-physical systems. VU-ISIS is a world-leading research institute in cyber-physical systems, and it is my honor to be part of it.

Extended from my current domain, I am interested in landing RL in real-world systems and making an impact on the society. This goal involves multiple challenges, including:

• How do humans involve in the RL loop?
• How RL can have long-term impacts on the society?
• What domains have the potential to be entirely reshaped by RL?

I received enormous help from our industrial collaborators, and I believe the industry is the best positioned to land modern technologies at scale and impact the society. Thus, I am always open to opportunities to connect, collaborate, and join the industry (across many domains!).

Some milestones of my research journey include (I use “we” to refer to myself and my collaborators since I always team up with others):

• We are the first to enable RL interacting with real-world commercial traffic signal controllers in real time.
• We are the first to introduce the concept of input-driven systems in the RL context to transportation domain and use it to formulation the traffic signal control problem.
• We published a traffic signal control simulation scenario with the longest temporal coverage of traffic volumes and with real signal timing data (3 years).
• We designed and deployed the world’s first RL-based variable speed limit (VSL) system on Interstate 24 southeast of Nashville, TN, USA.
• We conducted the world’s largest open road experiment using connected and automated vehicles (CAVs) to alleviate traffic congestion and smooth traffic flow.
• We won the CityLearn Challeng 2021, a multi-agent RL competition for urban energy management.

See my full publication record at Google Scholar.