Abstract: The scope of "smart city projects" aims at leveraging advanced technologies to improve the quality of living in modern cities. To achieve this goal, we need to gather real-time information from multiple sources, and integrate them for future state prediction as well as socially beneficial decision-making. This talk will discuss how to incorporate the state-of-the-art machine learning techniques into one crucial aspect in smart city applications, i.e., smart transportation. Specifically, we will introduce some recent academic works in three different levels: 1) On information acquisition level, we discuss how to use a "prophet" model combined with a random forest algorithm to impute missing traffic data to counter the issue of detector failures; 2) on prediction level, we discuss how to leverage a graph convolutional sequence-to-sequence model to capture the complex spatio-temporal dependencies in traffic networks and make multi-step traffic state predictions; 3) on decision-making level, we introduce a novel advisor-student reinforcement learning framework that combines machine learning and combinatorial optimization to tackle the dynamic operation of ride-hailing vehicle fleets. Some recent advancements of machine learning approaches applied in the field of connected and automated vehicles will also be briefly discussed.
Speaker Bio: Xi graduated from the Department of Civil Engineering, Tsinghua University. He is now a research associate at the Center for Sustainable Transportation at Tsinghua University, cooperating with a number of professors and co-supervising more than 20 PhD students. His research focuses on a series of engineering and management problems originated from innovative modes and advanced technological development in the transportation field, especially from the advancements of connected and automated vehicle technologies. He has published approximately 50 journal articles, mostly appeared on leading journals including Transportation Science, Transportation Research, IEEE Transactions on ITS, and Production and Operations Management, and he has won a number of awards as conference presenter and journal referee.